Ad hoc teamwork (AHT) requires agents to collaborate with previously unseen teammates, which is crucial for many real-world applications. The core challenge of AHT is to develop an ego agent that can predict and adapt to unknown teammates on the fly. Conventional RL-based approaches optimize a single expected return, which often causes policies to collapse into a single dominant behavior, thus failing to capture the multimodal cooperation patterns inherent in AHT. In this work, we introduce PADiff, a diffusion-based approach that captures agent's multimodal behaviors, unlocking its diverse cooperation modes with teammates. However, standard diffusion models lack the ability to predict and adapt in non-stationary AHT scenarios. To address this limitation, we propose a novel diffusion-based policy that integrates critical predictive information about teammates into the denoising process. Extensive experiments across three environments demonstrate that PADiff outperforms existing AHT methods significantly.

Offline Reinforcement Learning (RL) enables policy improvement from fixed datasets without online interactions, making it highly suitable for real-world applications lacking efficient simulators. Despite its success in the single-agent setting, offline multi-agent RL remains a challenge, especially in competitive games. Firstly, unaware of the game structure, it is impossible to interact with the opponents and conduct a major learning paradigm, self-play, for competitive games. Secondly, real-world datasets cannot cover all the state and action space in the game, resulting in barriers to identifying Nash equilibrium (NE). To address these issues, this paper introduces Off-FSP, the first practical model-free offline RL algorithm for competitive games. We start by simulating interactions with various opponents by adjusting the weights of the fixed dataset with importance sampling. This technique allows us to learn the best responses to different opponents and employ the Offline Self-Play learning framework. To overcome the challenge of partial coverage, we combine the single-agent offline RL method with Fictitious Self-Play (FSP) to approximate NE by constraining the approximate best responses away from out-of-distribution actions. Experiments on matrix games, extensive-form poker, and board games demonstrate that Off-FSP achieves significantly lower exploitability than state-of-the-art baselines. Finally, we validate Off-FSP on a real-world human-robot competitive task, demonstrating its potential for solving complex, hard-to-simulate real-world problems.

Lingyue Fu, Hao Guan, Bolun Zhang, Haowei Yuan, Yaoming Zhu, Lin Qiu, ZongYu Wang, Xuezhi Cao, Xunliang Cai, Weiwen Liu, Weinan Zhang, Yong Yu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Jianghao Lin, Yuanyuan Shi, Xin Peng, Renjie Ding, Hairui Wang, Yuxuan Peng, Bizhe Bai, Weixi Song, Fengshuo Bai, Huacan Chai, Weinan Zhang, Fei Huang, Ying Wen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Yunjia Xi, Jianghao Lin, Yongzhao Xiao, Zheli Zhou, Rong Shan, Te Gao, Jiachen Zhu, Weiwen Liu, Yong Yu, Weinan Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Zhiyuan Yao, Zishan Xu, Yifu Guo, Zhiguang Han, Cheng Yang, Shuo Zhang, Weinan Zhang, Xingshan Zeng, Weiwen Liu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Kangning Zhang, Weiwen Liu, Wenxiang Jiao, Kounianhua Du, Yuan Lu, Weinan Zhang, Yong Yu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Congmin Zheng, Jiachen Zhu, Zhuoying Ou, Yuxiang Chen, Kangning Zhang, Rong Shan, Zeyu Zheng, Mengyue Yang, Jianghao Lin, Yong Yu, Weinan Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

E-commerce has revolutionized retail, yet its traditional workflows remain inefficient, with significant resource costs tied to product design and inventory. This paper introduces a novel system deployed at Alibaba that uses AI-generated items (AIGI) to address these challenges with personalized text-to-image generation for e-commerce product design. AIGI enables an innovative business mode called "sell it before you make it", where merchants can design fashion items and generate photorealistic images with digital models based on textual descriptions. Only when the items have received a certain number of orders, do the merchants start to produce them, which largely reduces reliance on physical prototypes and thus accelerates time to market. For such a promising application, we identify the underlying key scientific challenge, i.e., capturing users' group-level personalized preferences towards multiple generated images. To this end, we propose a Personalized Group-Level Preference Alignment Framework for Diffusion Models (PerFusion). We first design PerFusion Reward Model for user preference estimation with a feature-crossing-based personalized plug-in. Then we develop PerFusion with a personalized adaptive network to model diverse preferences across users, and meanwhile derive the group-level preference optimization objective to model comparative behaviors among multiple images. Both offline and online experiments demonstrate the effectiveness of our proposed algorithm. The AI-generated items achieve over 13% relative improvements for both click-through rate and conversion rate, as well as 7.9% decrease in return rate, compared to their human-designed counterparts, validating the transformative potential of AIGI for e-commerce platforms.

Recently, large language models (LLMs) have advanced recommendation systems (RSs), and recent works have begun to explore how to integrate LLMs into industrial RSs. While most approaches deploy LLMs offline to generate and pre-cache augmented representations for RSs, high-dimensional representations from LLMs introduce substantial storage and computational costs. Thus, it is crucial to compress LLM representations effectively. However, we identify a counterintuitive phenomenon during representation compression: Mid-layer Representation Advantage (MRA), where representations from middle layers of LLMs outperform those from final layers in recommendation tasks. This degraded final layer renders existing compression methods, which typically compress on the final layer, suboptimal. We interpret this based on modularity theory that LLMs develop spontaneous internal functional modularity and force the final layer to specialize in the proxy training task. Thus, we propose Modular Representation Compression (MARC) to explicitly control the modularity of LLMs. First, Modular Adjustment explicitly introduces compression and task adaptation modules, enabling the LLM to operate strictly as a representation-learning module. Next, to ground each module to its specific task, Modular Task Decoupling uses information constraints and different network structures to decouple tasks. Extensive experiments validate that MARC addresses MRA and produces efficient representations. Notably, MARC achieved a 2.82% eCPM lift in an online A/B test within a large-scale commercial search advertising scenario.

The rapid advancement of multimodal large language models has enabled agents to operate mobile devices by directly interacting with graphical user interfaces, opening new possibilities for mobile automation. However, real-world mobile tasks are often complex and allow for multiple valid solutions. This contradicts current mobile agent evaluation standards: offline static benchmarks can only validate a single predefined ''golden path'', while online dynamic testing is constrained by the complexity and non-reproducibility of real devices, making both approaches inadequate for comprehensively assessing agent capabilities. To bridge the gap between offline and online evaluation and enhance testing stability, this paper introduces a novel graph-structured benchmarking framework. By modeling the finite states observed during real-device interactions, it achieves static simulation of dynamic behaviors. Building on this, we develop ColorBench, a benchmark focused on complex long-horizon tasks. It supports evaluation of multiple valid solutions, subtask completion rate statistics, and atomic-level capability analysis. ColorBench contains 175 tasks (74 single-app, 101 cross-app) with an average length of over 13 steps. Each task includes at least two correct paths and several typical error paths, enabling quasi-dynamic interaction.

In an era dominated by information overload, effective recommender systems are essential for managing the deluge of data across digital platforms. Multi-stage cascade ranking systems are widely used in the industry, with retrieval and ranking being two typical stages. Retrieval methods sift through vast candidates to filter out irrelevant items, while ranking methods prioritize these candidates to present the most relevant items to users. Unlike studies focusing on the ranking stage, this survey explores the critical yet often overlooked retrieval stage of recommender systems. To achieve precise and efficient personalized retrieval, we summarize existing work in three key areas: improving similarity computation between user and item, enhancing indexing mechanisms for efficient retrieval, and optimizing training methods of retrieval. We also provide a comprehensive set of benchmarking experiments on three public datasets. Furthermore, we highlight current industrial applications through a case study on retrieval practices at a specific company, covering the entire retrieval process and online serving, along with practical implications and challenges. By detailing the retrieval stage, which is fundamental for effective recommendation, this survey aims to bridge the existing knowledge gap and serve as a cornerstone for researchers interested in optimizing this critical component of cascade recommender systems.

As large language models (LLMs) achieve remarkable success in natural language processing (NLP) domains, LLM-enhanced recommender systems have received much attention and are being actively explored currently. In this article, we focus on adapting and enhancing large language models for recommendation tasks. First and foremost, we identify and formulate the lifelong sequential behavior incomprehension problem for LLMs in recommendation realms, i.e., LLMs fail to effectively extract useful information from a pure textual context of long user behavior sequence, even if the length of context is well below the context limitation of LLMs. To address such an issue and improve the recommendation performance of LLMs, we propose a novel framework, namely, R etrieval- e nhanced L arge La nguage models Plus (ReLLaX), which provides full-stack optimization from three perspectives, i.e., data, prompt, and parameter. For data-level enhancement, we design semantic user behavior retrieval (SUBR) to reduce the heterogeneity of the behavior sequence, thus lowering the difficulty for LLMs to extract the essential information from user behavior sequences. Although SUBR can improve the data quality, further increase in the sequence length will still raise its heterogeneity to a level where LLMs can no longer comprehend it. Hence, we further propose to perform prompt-level and parameter-level enhancement, with the integration of conventional recommendation models (CRMs). As for prompt-level enhancement, we apply soft prompt augmentation (SPA) to explicitly inject collaborative knowledge from CRMs into the prompt. The item representations of LLMs are thus more aligned with recommendation, helping LLMs better explore the item relationships in the sequence and facilitating comprehension. Finally, for parameter-level enhancement, we propose component fully-interactive LoRA (CFLoRA). By enabling sufficient interaction between the LoRA atom components, the expressive ability of LoRA is extended, making the parameters effectively capture more sequence information. Moreover, we present new perspectives to compare current LoRA-based LLM4Rec methods, i.e., from both a composite and a decomposed view. We theoretically demonstrate that the ways they employ LoRA for recommendation are degraded versions of our CFLoRA, with different constraints on atom component interactions. Extensive experiments are conducted on three real-world public datasets to demonstrate the superiority of ReLLaX compared with existing baseline models, as well as its capability to alleviate lifelong sequential behavior incomprehension. Our code is available. 1

Recommender system plays a pervasive role in today’s online services, yet its closed-loop nature, i.e., training and deploying within a specific closed domain, constrains its access to open-world knowledge. Recently, the emergence of large language models (LLMs) has shown promise in bridging this gap by encoding extensive world knowledge and demonstrating advanced reasoning capabilities. However, previous attempts to directly implement LLMs as recommenders fall short in meeting the demanding requirements of industrial recommender systems, particularly in terms of online inference latency and offline resource efficiency. In this work, we propose an Open-World R ecommendation Framework with E fficient and Deployable K nowledge I nfusion from Large Language Models, dubbed REKI , to acquire two types of external knowledge about users and items from LLMs. Specifically, we introduce factorization prompting to elicit accurate knowledge reasoning on user preferences and items. With factorization prompting, we develop individual knowledge extraction and collective knowledge extraction tailored for different scales of recommendation scenarios, effectively reducing offline resource consumption. Subsequently, the generated user and item knowledge undergoes efficient transformation and condensation into augmented vectors through a hybridized expert-integrated network , ensuring its compatibility with the recommendation task. The obtained vectors can then be directly used to enhance the performance of any conventional recommendation model. We also ensure efficient inference by preprocessing and prestoring the knowledge from the LLM. Extensive experiments demonstrate that REKI significantly outperforms the state-of-the-art baselines and is compatible with a diverse array of recommendation algorithms and tasks. Now, REKI has been deployed to Huawei’s news and music recommendation platforms and gained a 7% and 1.99% improvement during the online A/B test.

Given the large volume of side information from different modalities, multimodal recommender systems have become increasingly vital, as they exploit richer semantic information beyond user-item interactions. Recent works highlight that leveraging Graph Convolutional Networks (GCNs) to explicitly model multimodal item-item relations can significantly enhance recommendation performance. However, due to the inherent over-smoothing issue of GCNs, existing models benefit only from shallow GCNs with limited representation power. This drawback is especially pronounced when facing complex and high-dimensional patterns such as multimodal data, as it requires large-capacity models to accommodate complicated correlations. To this end, in this paper, we investigate bypassing GCNs when modeling multimodal item-item relationship. More specifically, we propose a Topology-aware Multi-Layer Perceptron (TMLP), which uses MLPs instead of GCNs to model the relationships between items. TMLP enhances MLPs with topological pruning to denoise item-item relations and intra (inter)-modality learning to integrate higher-order modality correlations. Extensive experiments on three real-world datasets verify TMLP's superiority over nine baselines. We also find that by discarding the internal message passing in GCNs, which is sensitive to node connections, TMLP achieves significant improvements in both training efficiency and robustness against existing models.

Reinforcement Learning has revolutionized decision-making processes in dynamic environments, yet it often struggles with autonomously detecting and achieving goals without clear feedback signals. For example, in a Source Term Estimation problem, the lack of precise environmental information makes it challenging to provide clear feedback signals and to define and evaluate how the source's location is determined. To address this challenge, the Autonomous Goal Detection and Cessation (AGDC) module was developed, enhancing various RL algorithms by incorporating a self-feedback mechanism for autonomous goal detection and cessation upon task completion. Our method effectively identifies and ceases undefined goals by approximating the agent's belief, significantly enhancing the capabilities of RL algorithms in environments with limited feedback. To validate effectiveness of our approach, we integrated AGDC with deep Q-Network, proximal policy optimization, and deep deterministic policy gradient algorithms, and evaluated its performance on the Source Term Estimation problem. The experimental results showed that AGDC-enhanced RL algorithms significantly outperformed traditional statistical methods such as infotaxis, entrotaxis, and dual control for exploitation and exploration, as well as a non-statistical random action selection method. These improvements were evident in terms of success rate, mean traveled distance, and search time, highlighting AGDC's effectiveness and efficiency in complex, real-world scenarios.

With the impressive reasoning and text generation capabilities of large language models (LLMs), methods leveraging multiple LLMs to debate each other have garnered increasing attention. However, existing debate-based approaches remain limited in effectiveness in structured and detailed domains represented by code generation due to several reasons: 1) Reliance on different instances of the same LLM for debate, neglecting the potential benefits of integrating diverse models with varied internal knowledge for more comprehensive code generation, 2) under-utilization of test cases, and 3) reliance on third-party LLM moderators for result consolidation and decision-making, probably introducing hallucinations and judgment errors. To address these challenges, we propose DebateCoder to collect intelligence of LLMs via test case-driven debate for code generation. In DebateCoder, test cases serve as a medium for models to analyze code and identify bugs, while opposing models generate test cases to challenge each other’s code during the debate process. These test cases, along with their execution results, are elaborately leveraged to refine and enhance the code through a novel contrastive analysis process. Furthermore, DebateCoder leverages test case outcomes to assess code quality and determine convergence criteria. Unlike previous approaches, DebateCoder emphasizes the collaborative improvement of both models through competitive debate and interactive analysis. Abundant experimental results on two datasets demonstrate the effectiveness of DebateCoder.

Shao Zhang, Xihuai Wang, Wenhao Zhang, Chaoran Li, Junru Song, Tingyu Li, Lin Qiu, Xuezhi Cao, Xunliang Cai, Wen Yao, Weinan Zhang, Xinbing Wang, Ying Wen. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Cognitive diagnosis (CD) plays a crucial role in intelligent education, evaluating students' comprehension of knowledge concepts based on their test histories. However, current CD methods often model students, exercises, and knowledge concepts solely on their ID relationships, neglecting the abundant semantic relationships present within the educational data space. Furthermore, contemporary intelligent tutoring systems (ITS) frequently involve the addition of new students and exercises, creating cold-start scenarios that ID-based methods find challenging to manage effectively. The advent of large language models (LLMs) offers the potential for overcoming this challenge with open-world knowledge. In this paper, we propose LLM4CD, which Leverages Large Language Models for open-world knowledge Augmented Cognitive Diagnosis. Our method utilizes the open-world knowledge of LLMs to construct cognitively expressive textual representations, which are then encoded to introduce rich semantic information into the CD task. Additionally, we propose an innovative bi-level encoder framework that models students' test histories through two levels of encoders: a macro-level cognitive text encoder and a micro-level knowledge state encoder. This approach substitutes traditional ID embeddings with semantic representations, enabling the model to accommodate new students and exercises with open-world knowledge and address the cold-start problem. Extensive experimental results demonstrate that LLM4CD consistently outperforms previous CD models on multiple real-world datasets, validating the effectiveness of leveraging LLMs to introduce rich semantic information into the CD task.

Recent advances in embodied intelligence have opened new directions for autonomous agents to operate in complex physical environments. In this work, we propose a Posture-Aware Locomotion Agent (PALA) trained via deep reinforcement learning. Unlike traditional quadruped agents that focus solely on velocity tracking, our agent learns to track task-oriented 6D motion commands, including linear and angular velocities, as well as desired body posture (height, pitch, and roll), in real time using only proprioceptive sensing and external commands. To improve robustness and terrain adaptability, we introduce two key heuristic designs: a progressive reward curriculum and an orientation command resampling strategy. Combined with asymmetric actor-critic training, adversarial motion priors, and domain randomization, these components enable a single policy to generalize zero-shot across diverse and challenging environments. Additionally, we extend PALA to handle high-level instructions by integrating it with an autonomous agent powered by a large language model (LLM), enabling natural language task descriptions to be directly translated into executable 6D commands. Extensive simulation and real-world experiments demonstrate the responsiveness and accuracy of our posture-aware locomotion agent, underscoring its potential as a core component for embodied systems operating in unstructured settings.

Fine-tuning large language models (LLMs) is challenged by the presence of noisy data and the high computational cost when training on large-scale datasets. While data selection has emerged as a promising approach to reduce training cost and improve data quality, existing methods often rely on static heuristics or manual metrics. These approaches struggle to adapt to the model’s evolving capabilities during training, as its understanding of tasks improves. As the model becomes more powerful, its requirements for data that can enhance performance also change, making it crucial to incorporate this dynamic into the data selection process. Moreover, ensuring data diversity throughout different stages of training is essential for preventing redundancy, reducing overfitting. To address these issues, we propose DSP, a Diversity-Aware Self-Paced data selection framework that evolves with the model. DSP progressively selects training samples based on the model’s own outputs and incorporates a diversity-aware mechanism to enhance generalization and mitigate overfitting. Unlike prior static or rule-based strategies, DSP adaptively adjusts to the model’s internal feedback and training stage. Experiments on two public benchmarks demonstrate that DSP consistently outperforms static and heuristic-based baselines across multiple datasets and backbone models. Our findings highlight the critical role of dynamic, diversity-aware data selection in effective LLM fine-tuning.

Qingyao Li, Wei Xia, Xinyi Dai, Kounianhua Du, Weiwen Liu, Yasheng Wang, Ruiming Tang, Yong Yu, Weinan Zhang. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Weiming Zhang, Qingyao Li, Xinyi Dai, Jizheng Chen, Kounianhua Du, Weiwen Liu, Yasheng Wang, Ruiming Tang, Yong Yu, Weinan Zhang. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

An important challenge in multi-agent reinforcement learning is partial observability, where agents cannot access the global state of the environment during execution and can only receive observations within their field of view. To address this issue, previous works typically use the dimensional-wise state, which is obtained by applying MLP or dimensional-based attention on the global state, for decision-making during training and relying on a reconstructed dimensional-wise state during execution. However, dimensional-wise states tend to divert agent attention to specific features, neglecting potential dependencies between agents, making it difficult to make optimal decisions. Moreover, the inconsistency between the states used in training and execution further increases additional errors. To resolve these issues, we propose a method called Reconstruction-Guided Policy (RGP) to reconstruct the agent-wise state, which represents the information of inter-agent relationships, as input for decision-making during both training and execution. This not only preserves the potential dependencies between agents but also ensures consistency between the states used in training and execution. We conducted extensive experiments on both discrete and continuous action environments to evaluate RGP, and the results demonstrates its superior effectiveness. Our code is public in https://anonymous.4open.science/r/RGP-9F79

Large language models have demonstrated impressive value in performing as autonomous agents when equipped with external tools and API calls. Nonetheless, effectively harnessing their potential for executing complex tasks crucially relies on enhancements in their function-calling capabilities. This paper identifies a critical gap in existing function-calling models, where performance varies significantly across benchmarks, often due to over-fitting to specific naming conventions. To address such an issue, we introduce Hammer, a novel family of foundation models specifically engineered for on-device function calling. Hammer employs an augmented dataset that enhances models’ sensitivity to irrelevant functions and incorporates function masking techniques to minimize over-fitting. Our empirical evaluations reveal that Hammer not only outperforms larger models but also demonstrates robust generalization across diverse benchmarks, achieving state-of-the-art results. Our open-source contributions include a specialized dataset for irrelevance detection, a tuning framework for enhanced generalization, and the Hammer models, establishing a new standard for function-calling performance.

The performance of offline reinforcement learning (RL) is sensitive to the proportion of high-return trajectories in the offline dataset. However, in many simulation environments and real-world scenarios, there are large ratios of low-return trajectories rather than high-return trajectories, which makes learning an efficient policy challenging. In this paper, we propose a method called Contrastive Diffuser (ContraDiff) to make full use of low-return trajectories and improve the performance of offline RL algorithms. Specifically, ContraDiff groups the states of trajectories in the offline dataset into high-return states and low-return states and treats them as positive and negative samples correspondingly. Then, it designs a contrastive mechanism to pull the planned trajectory of an agent toward high-return states and push them away from low-return states. Through the contrast mechanism, trajectories with low returns can serve as negative examples for policy learning, guiding the agent to avoid areas associated with low returns and achieve better performance. Through the contrast mechanism, trajectories with low returns provide a ``counteracting force'' guides the agent to avoid areas associated with low returns and achieve better performance. Experiments on 27 sub-optimal datasets demonstrate the effectiveness of our proposed method. Our code is publicly available at https://github.com/Looomo/contradiff.

In-context learning with large language models (LLMs) delivers strong few-shot performance by choosing few-shot demonstrations from the entire training dataset. However, previous few-shot in-context learning methods, which calculate similarity scores for choosing demonstrations, incur high computational costs by repeatedly retrieving large-scale datasets for each query. This is due to their failure to recognize that not all demonstrations are equally informative, and many less informative demonstrations can be inferred from a core set of highly informative ones. To this end, we propose FEEDER (FEw yet Essential Demonstration prE-selectoR), a novel \emph{pre-selection} framework that identifies a core subset of demonstrations containing the most informative examples. This subset, referred to as the FEEDER set, consists of demonstrations that capture both the ''sufficiency'' and ''necessity'' information to infer the entire dataset. Notice that FEEDER is selected before the few-shot in-context learning, enabling more efficient few-shot demonstrations choosing in a smaller set. To identify FEEDER, we propose a novel effective tree based algorithm. Once selected, it can replace the original dataset, leading to improved efficiency and prediction accuracy in few-shot in-context learning. Additionally, FEEDER also benefit fine-tuning LLMs, we propose a bi-level optimization method enabling more efficient training without sacrificing performance when datasets become smaller. Our experiments are on 6 text classification datasets, 1 reasoning dataset, and 1 semantic-parsing dataset, across 6 LLMs (ranging from 335M to 7B parameters), demonstrate that: (i) In few-shot inference, FEEDER achieves superior (or comparable) performance while utilizing only half the input training data. (ii) In fine-tuning, FEEDER significantly boosts the performance of LLMs.

Diffusion policies have shown promise in learning complex behaviors from demonstrations, particularly for tasks requiring precise control and long-term planning. However, they face challenges in robustness when encountering distribution shifts. This paper explores improving diffusion-based imitation learning models through online interactions with the environment. We propose OTPR (Optimal Transport-guided score-based diffusion Policy for Reinforcement learning fine-tuning), a novel method that integrates diffusion policies with RL using optimal transport theory. OTPR leverages the Q-function as a transport cost and views the policy as an optimal transport map, enabling efficient and stable fine-tuning. Moreover, we introduce masked optimal transport to guide state-action matching using expert keypoints and a compatibility-based resampling strategy to enhance training stability. Experiments on three simulation tasks demonstrate OTPR’s superior performance and robustness compared to existing methods, especially in complex and sparse-reward environments. In sum, OTPR provides an effective framework for combining IL and RL, achieving versatile and reliable policy learning.

Legged locomotion over various terrains is challenging and requires precise perception of the robot and its surroundings from both proprioception and vision. However, learning directly from high-dimensional visual input is often data-inefficient and intricate. To address this issue, traditional methods attempt to learn a teacher policy with access to privileged information first and then learn a student policy to imitate the teacher's behavior with visual input. Despite some progress, this imitation framework prevents the student policy from achieving optimal performance due to the information gap between inputs. Furthermore, the learning process is unnatural since animals intuitively learn to traverse different terrains based on their understanding of the world without privileged knowledge. Inspired by this natural ability, we propose a simple yet effective method, World Model-based Perception (WMP), which builds a world model of the environment and learns a policy based on the world model. We illustrate that though completely trained in simulation, the world model can make accurate predictions of real-world trajectories, thus providing informative signals for the policy controller. Extensive simulated and real-world experiments demonstrate that WMP outperforms state-of-the-art baselines in traversability and robustness. Videos and Code are available at: https://wmp-loco.github.io/.

Second-order Online Kernel Learning (OKL) has attracted considerable research interest due to its promising predictive performance in streaming environments. However, existing second-order OKL approaches suffer from at least quadratic time complexity with respect to the pre-set budget, rendering them unsuitable for large-scale datasets. Moreover, the singular value decomposition required to obtain explicit feature mapping is computationally expensive due to the complete decomposition process. To address these issues, we propose FORKS, a fast incremental matrix sketching and decomposition approach tailored for second-order OKL. FORKS constructs an incremental maintenance paradigm for second-order kernelized gradient descent, which includes incremental matrix sketching for kernel approximation and incremental matrix decomposition for explicit feature mapping construction. Theoretical analysis demonstrates that FORKS achieves a logarithmic regret guarantee on par with other second-order approaches while maintaining a linear time complexity w.r.t. the budget, significantly enhancing efficiency over existing methods. We validate the performance of our method through extensive experiments conducted on real-world datasets, demonstrating its superior scalability and robustness against adversarial attacks.

The goal of offline model-based optimization (MBO) is to seek designs that maximize a black-box function with only offline historical function evaluations, which has a wide range of applications. Since the black-box function cannot be queried directly, a surrogate model is usually learned to approximate the objective function. To overcome the challenge of distribution shift between the empirical offline data distribution and the underlying data distribution, the literature typically applies regularization to the surrogate. In this work, we propose Fisher Regularized Model (FiRM), a novel forward approach to offline MBO. FiRM decomposes a forward surrogate into an offline density model and an offset network, and applies gradient regularization on the offset to constrain the optimized designs to stay close to the offline data support, so as to make a proper trade-off between conservative estimation and generalization. In addition, from the perspective of energy-based model (EBM), we prove that the gradient regularization is equivalent to a Fisher-divergence constraint to the surrogate. Further theoretical analysis establishes connections between FiRM and existing work. Experimental results on five different tasks in a standard benchmark show that our method outperforms all the baselines, achieving the highest mean score and mean rank.

Model merging has gained increasing attention due to its intriguing property: interpolating the parameters of different task-specific fine-tuned models leads to multi-task abilities. However, despite its empirical success, the underlying mechanisms of model merging remain poorly understood. In this work, we delve into the mechanism behind model merging from a representation perspective. Our analysis reveals that model merging achieves multi-task abilities through two key capabilities: i) distinguishing samples from different tasks, and ii) adapting to the corresponding expert model for each sample. These two capabilities allow the merged model to retain task-specific expertise, enabling efficient multi-task adaptation. Building on these insights, we propose SE-Merging, a self-enhanced model merging framework that leverages these two characteristics to dynamically identify the corresponding task for each sample and then adaptively rescales the merging coefficients to further enhance task-specific expertise in the merged model. Notably, SE-Merging achieves dynamic model merging without additional training. Extensive experiments demonstrate that SE-Merging achieves significant performance improvements while remaining compatible with existing model merging techniques.

Computerized Adaptive Testing (CAT) is a fundamental issue in intelligent education, which aims to select a relatively small number of questions to assess student’s ability. However, existing approaches often suffer from a limited selection area within the question bank, as questions without ground truth answers in the practice log are considered invalid for selection. The selection area refers to the range of questions in the question bank that have answers. To expand the selection area, we proposed a framework called Certainty-Guaranteed Environment for Computerized Adaptive Testing (CECAT). First, Simulative Interaction Module (SIM) leverages questions and their answer results in practice log in training data, so as to predict the distribution of answer result for each question without answer result in practice log. Second, α-Certainty is designed to measure the statistical similarity of the predicted distribution of answer results obtained from SIM, so as to exclude the statistically dissimilar expanded answer results. Extensive experiments on three public datasets demonstrate that CECAT outperforms twelve strong baselines in terms of assessing the student’s ability. Our code is available at https://anonymous.4open.science/r/SimCAT-F31B/README.md.

Computerized Adaptive Testing (CAT) is a burgeoning online educational technology that assesses student ability through a relatively small number of carefully selected questions. Assessing a student’s ability from his/her short question-answer sequence is a few-shot learning problem, which is currently solved by meta-learning-based methods with a bi-level structure. However, current methods using the bi-level structure have limitations, such as inefficiency due to alternating optimization between the inner level and outer level and not fully using historical students. To address these, we propose the Individualized High-Efficiency Ability Assessment for Computerized Adaptive Testing (IECAT), which introduces (1) a Student Comparison method to fully use dissimilar students by distinguishing similar and dissimilar students, (2) an Ability Matrix to fully use practice logs of historical students to get more accurate abilities, and (3) an Individual Mapping (IM) structure to improve efficiency by optimizing in its single-level structure. Extensive experiments across three public datasets demonstrate that IECAT outperforms twelve strong baselines. Furthermore, our experiments validate that IM structure consumes less training time than bilevel structure. Our code is available at https://anonymous.4open.science/r/IECAT-8FE5/IECAT/ReadMe.txt.

Reinforcement Learning (RL) has shown its remarkable and generalizable capability in legged locomotion through sim-to-real transfer. However, while adaptive methods like domain randomization are expected to enhance policy robustness across diverse environments, they potentially compromise the policy’s performance in any specific environment, leading to suboptimal real-world deployment due to the No Free Lunch Theorem. To address this, we propose LoopSR, a lifelong policy adaptation framework that continuously refines RL policies in the post-deployment stage. LoopSR employs a transformer-based encoder to map real-world trajectories into a latent space and reconstruct a digital twin of the real world for further improvement. Autoencoder architecture and contrastive learning methods are adopted to enhance feature extraction of real-world dynamics. Simulation parameters for continual training are derived by combining predicted values from the decoder with retrieved parameters from a pre-collected simulation trajectory dataset. By leveraging simulated continual training, LoopSR achieves superior data efficiency compared with strong baselines, yielding eminent performance with limited data in both sim-to-sim and sim-to-real experiments.

Humans possess delicate dynamic balance mechanisms that enable them to maintain stability across diverse terrains and under extreme conditions. However, despite significant advances recently, existing locomotion algorithms for humanoid robots are still struggle to traverse extreme environments, especially in cases that lack external perception (e.g., vision or LiDAR). This is because current methods often rely on gait-based or perception-condition rewards, lacking effective mechanisms to handle unobservable obstacles and sudden balance loss. To address this challenge, we propose a novel whole-body locomotion algorithm based on dynamic balance and Reinforcement Learning (RL) that enables humanoid robots to traverse extreme terrains, particularly narrow pathways and unexpected obstacles, using only proprioception. Specifically, we introduce a dynamic balance mechanism by leveraging a novel Zero Moment Point (ZMP)-driven reward and task-driven rewards in a whole-body actor-critic framework, aiming to achieve coordinated actions of the upper and lower limbs for robust locomotion. Experiments conducted on a full-sized Unitree H1-2 robot verify the ability of our method to maintain balance on extremely narrow terrains and under external disturbances, demonstrating its effectiveness in enhancing the robot's adaptability to complex environments. The videos are given at https://whole-body-loco.github.io.

Microscopic traffic simulation has become an important tool for autonomous driving training and testing. Although recent data-driven approaches advance realistic behavior generation, their learning still relies primarily on a single real-world dataset, which limits their diversity and thereby hinders downstream algorithm optimization. In this paper, we propose DriveGen, a novel traffic simulation framework with large models for more diverse traffic generation that supports further customized designs. DriveGen consists of two internal stages: the initialization stage uses a large language model and retrieval technique to generate map and vehicle assets; the rollout stage outputs trajectories with selected waypoint goals from a visual language model and a specifically designed diffusion planner. Through this two-staged process, DriveGen fully utilizes large models’ high-level cognition and reasoning of driving behavior, obtaining greater diversity beyond datasets while maintaining high realism. To support effective downstream optimization, we additionally develop DriveGen-CS, an automatic corner case generation pipeline that uses failures of the driving algorithm as additional prompt knowledge for large models without the need for retraining or fine-tuning. Experiments show that our generated scenarios and corner cases have superior performance compared to state-of-the-art baselines. Downstream experiments further verify that the synthesized traffic of DriveGen provides better optimization of the performance of typical driving algorithms, demonstrating the effectiveness of our framework.

Recommender systems (RSs) have become essential for alleviating information overload and matching users with relevant content.Traditionally, RSs have focused on personalized content distribution, leveraging user interaction data and various features to rank and recommend existing items.Recently, diffusion models (DMs) have emerged as powerful generative paradigms, introducing new possibilities for RSs to not only enhance their performance for content distribution but also extend their capability boundaries to personalized content creation.On the one hand, DMs enhance the recommendation performance by mitigating challenges such as sparse user-item interactions, weak latent representations, and noisy data.On the other hand, DMs enable personalized content creation, transforming RSs from passive distributors into active generators of user-specific media assets, such as customized images, posters, and multimedia content.Given such a transformative paradigm shift, this survey provides a comprehensive review of the integration of diffusion models into recommender systems, exploring key methodologies, application scenarios, and their impact on recommendation effectiveness, diversity, and personalization.We categorize DM-based recommendation paradigms into content distribution and content creation, compare integration strategies, and discuss open challenges and future directions.This work aims to guide researchers and practitioners in developing the next generation of generative AI-powered recommendation solutions.

Graph neural networks (GNNs) have emerged as state-of-the-art methods to learn from graph-structured data for recommendation. However, most existing GNN-based recommendation methods focus on the optimization of model structures and learning strategies based on pre-defined graphs, neglecting the importance of the graph construction stage. Earlier works for graph construction usually rely on specific rules or crowdsourcing, which are either too simplistic or too labor-intensive. Recent works start to utilize large language models (LLMs) to automate the graph construction, in view of their abundant open-world knowledge and remarkable reasoning capabilities. Nevertheless, they generally suffer from two limitations: (1) invisibility of global view (e.g., overlooking contextual information) and (2) construction inefficiency. To this end, we introduce AutoGraph, an automatic graph construction framework based on LLMs for recommendation. Specifically, we first use LLMs to infer the user preference and item knowledge, which is encoded as semantic vectors. Next, we employ vector quantization to extract the latent factors from the semantic vectors. The latent factors are then incorporated as extra nodes to link the user/item nodes, resulting in a graph with in-depth global-view semantics. We further design metapath-based message aggregation to effectively aggregate the semantic and collaborative information. The framework is model-agnostic and compatible with different backbone models. Extensive experiments on three real-world datasets demonstrate the efficacy and efficiency of AutoGraph compared to existing baseline methods. We have deployed AutoGraph in Huawei advertising platform, and gain a 2.69% improvement on RPM and a 7.31% improvement on eCPM in the online A/B test. Currently AutoGraph has been used as the main traffic model, serving hundreds of millions of people.

Recommender systems (RSs) often suffer from the feedback loop phenomenon, i.e., RSs are trained on data biased by their recommendations. This leads to the filter bubble effect that reinforces homogeneous content and reduces user satisfaction. To this end, serendipity recommendations, which offer unexpected yet relevant items, are proposed. Recently, large language models (LLMs) have shown potential in serendipity prediction due to their extensive world knowledge and reasoning capabilities. However, they still face challenges in aligning serendipity judgments with human assessments, handling long user behavior sequences, and meeting the latency requirements of industrial RSs. To address these issues, we propose SERAL (Serendipity Recommendations with Aligned Large Language Models), a framework comprising three stages: (1) Cognition Profile Generation to compress user behavior into multi-level profiles; (2) SerenGPT Alignment to align serendipity judgments with human preferences using enriched training data; and (3) Nearline Adaptation to integrate SerenGPT into industrial RSs pipelines efficiently. Online experiments demonstrate that SERAL improves exposure ratio (PVR), clicks, and transactions of serendipitous items by 5.7%, 29.56%, and 27.6%, enhancing user experience without much impact on overall revenue. Now, it has been fully deployed in ''Guess What You Like'' on the Taobao App homepage.

Recent advances in reinforcement learning (RL) have demonstrated the powerful exploration capabilities and multimodality of generative diffusion-based policies. While substantial progress has been made in offline RL and off-policy RL settings, integrating diffusion policies into on-policy frameworks like PPO remains underexplored. This gap is particularly significant given the widespread use of large-scale parallel GPU-accelerated simulators, such as IsaacLab, which are optimized for on-policy RL algorithms and enable rapid training of complex robotic tasks. A key challenge lies in computing state-action log-likelihoods under diffusion policies, which is straightforward for Gaussian policies but intractable for flow-based models due to irreversible forward-reverse processes and discretization errors (e.g., Euler-Maruyama approximations). To bridge this gap, we propose GenPO, a generative policy optimization framework that leverages exact diffusion inversion to construct invertible action mappings. GenPO introduces a novel doubled dummy action mechanism that enables invertibility via alternating updates, resolving log-likelihood computation barriers. Furthermore, we also use the action log-likelihood for unbiased entropy and KL divergence estimation, enabling KL-adaptive learning rates and entropy regularization in on-policy updates. Extensive experiments on eight IsaacLab benchmarks, including legged locomotion (Ant, Humanoid, Anymal-D, Unitree H1, Go2), dexterous manipulation (Shadow Hand), aerial control (Quadcopter), and robotic arm tasks (Franka), demonstrate GenPO’s superiority over existing RL baselines. Notably, GenPO is the first method to successfully integrate diffusion policies into on-policy RL, unlocking their potential for large-scale parallelized training and real-world robotic deployment.

Evaluating large language models (LLMs) poses significant challenges, particularly due to issues of data contamination and the leakage of correct answers. To address these challenges, we introduce ThinkBench, a novel evaluation framework designed to robustly evaluate the reasoning capability of LLMs. ThinkBench proposes a dynamic data generation method for constructing out-of-distribution (OOD) datasets and offers an OOD dataset that contains 2,912 samples drawn from reasoning tasks. ThinkBench unifies the evaluation of reasoning models and non-reasoning models. We evaluate 16 LLMs and 4 PRMs under identical experimental conditions and show that most of the LLMs' performance are far from robust and they face a certain level of data leakage. By dynamically generating OOD datasets, ThinkBench effectively provides a reliable evaluation of LLMs and reduces data contamination impact. Our data and codes are available at https://github.com/huangshulin123/ThinkBench.

Recent advances in Multimodal Large Language Models (MLLMs) have enabled the development of mobile agents that can understand visual inputs and follow user instructions, unlocking new possibilities for automating complex tasks on mobile devices. However, applying these models to real-world mobile scenarios remains a significant challenge due to the long-horizon task execution, difficulty in error recovery, and the cold-start problem in unfamiliar environments. To address these challenges, we propose MobileUse, a GUI agent designed for robust and adaptive mobile task execution. To improve resilience in long-horizon tasks and dynamic environments, we introduce a hierarchical reflection architecture that enables the agent to self-monitor, detect, and recover from errors across multiple temporal scales—ranging from individual actions to overall task completion—while maintaining efficiency through a Reflection-on-Demand strategy. To tackle cold-start issues, we further introduce a proactive exploration module, which enriches the agent’s understanding of the environment through self-planned exploration. Evaluations on the AndroidWorld and AndroidLab benchmarks demonstrate that MobileUse establishes new state-of-the-art performance, achieving success rates of 62.9% and 44.2%, respectively. To facilitate real-world applications, we release an out-of-the-box toolkit for automated task execution on physical mobile devices, which is available at https://github.com/MadeAgents/mobile-use.

Survey papers are foundational to the scholarly progress of research communities, offering structured overviews that guide both novices and experts across disciplines. However, the recent surge of AI-generated surveys, especially enabled by large language models (LLMs), has transformed this traditionally labor-intensive genre into a low-effort, high-volume output. While such automation lowers entry barriers, it also introduces a critical threat: the phenomenon we term the "survey paper DDoS attack" to the research community. This refers to the unchecked proliferation of superficially comprehensive but often redundant, low-quality, or even hallucinated survey manuscripts, which floods preprint platforms, overwhelms researchers, and erodes trust in the scientific record. In this position paper, we argue that we must stop uploading massive amounts of AI-generated survey papers (i.e., survey paper DDoS attack) to the research community, by instituting strong norms for AI-assisted review writing. We call for restoring expert oversight and transparency in AI usage and, moreover, developing new infrastructures such as Dynamic Live Surveys, community-maintained, version-controlled repositories that blend automated updates with human curation. Through quantitative trend analysis, quality audits, and cultural impact discussion, we show that safeguarding the integrity of surveys is no longer optional but imperative to the research community.

Obtaining a generalizable reward model is crucial in Reinforcement Learning from Human Feedback (RLHF) as it enables correctly evaluating unseen prompt-response pairs. However, existing reward models lack this ability, as they are typically trained by increasing the reward gap between chosen and rejected responses, while overlooking the prompts that the responses are conditioned on. Consequently, when the trained reward model is evaluated on prompt-response pairs that lie outside the data distribution, neglecting the effect of prompts may result in poor generalization of the reward model. To address this issue, we decompose the reward value into two independent components: prompt-free reward and prompt-related reward. Prompt-free reward represents the evaluation that is determined only by responses, while the prompt-related reward reflects the reward that derives from both the prompt and the response. We extract these two components from an information theoretical perspective, which requires no extra models.. Subsequently, we propose a new reward learning algorithm by prioritizing data samples based on their prompt-free reward values. Through toy examples, we demonstrate that the extracted prompt-free and prompt-related rewards effectively characterize two parts of the reward model. Further, standard evaluations show that our method improves both the alignment performance and the generalization capability of the reward model.

Recent research on Reasoning of Large Language Models (LLMs) has sought to further enhance their performance by integrating meta-thinking—enabling models to monitor, evaluate, and control their reasoning processes for more adaptive and effective problem-solving. However, current single-agent work lacks a specialized design for acquiring meta-thinking, resulting in low efficacy. To address this challenge, we introduce Reinforced Meta-thinking Agents (ReMA), a novel framework that leverages Multi-Agent Reinforcement Learning (MARL) to elicit meta-thinking behaviors, encouraging LLMs to think about thinking. ReMA decouples the reasoning process into two hierarchical agents: a high-level meta-thinking agent responsible for generating strategic oversight and plans, and a low-level reasoning agent for detailed executions. Through iterative reinforcement learning with aligned objectives, these agents explore and learn collaboration, leading to improved generalization and robustness. Empirical results from single-turn experiments demonstrate that ReMA outperforms single-agent RL baselines on complex reasoning tasks, including competitive-level mathematical benchmarks and LLM-as-a-Judge benchmarks. Additionally, we further extend ReMA to multi-turn interaction settings, leveraging turn-level ratio and parameter sharing to improve efficiency. Comprehensive ablation studies further illustrate the evolving dynamics of each distinct agent, providing valuable insights into how the meta-thinking reasoning process enhances the reasoning capabilities of LLMs.

Humanoid robots are promising to acquire various skills by imitating human behaviors. However, existing algorithms are only capable of tracking smooth, low-speed human motions, even with delicate reward and curriculum design. This paper presents a physics-based humanoid control framework, aiming to master highly-dynamic human behaviors such as Kungfu and dancing through multi-steps motion processing and adaptive motion tracking. For motion processing, we design a pipeline to extract, filter out, correct, and retarget motions, while ensuring compliance with physical constraints to the maximum extent. For motion imitation, we formulate a bi-level optimization problem to dynamically adjust the tracking accuracy tolerance based on the current tracking error, creating an adaptive curriculum mechanism. We further construct an asymmetric actor-critic framework for policy training. In experiments, we train whole-body control policies to imitate a set of highly dynamic motions. Our method achieves significantly lower tracking errors than existing approaches and is successfully deployed on the Unitree G1 robot, demonstrating stable and expressive behaviors. The project page is https://kungfubot.github.io.

The practical use of reinforcement learning (RL) requires handling diverse settings, including online, offline, and offline-to-online learning. Instead of developing separate algorithms for each setting, we propose Uni-RL, a unified model-free RL framework that addresses all these scenarios within a single formulation. Uni-RL builds on the Implicit Value Regularization (IVR) framework and generalizes its dataset behavior constraint to the constraint w.r.t a reference policy, yielding an unified value learning objective for general settings. The reference policy is chosen to be the target policy in the online setting and the behavior policy in the offline setting. Using an iteratively refined behavior policy solves the over-constrained problem of directly applying IVR in the online setting, it provides an implicit trust-region style update through the value function while being off-policy. Uni-RL also introduces an unified policy extraction objective that estimates in-sample policy gradient using only actions from the reference policy. This supports various policy classes and theoretically guaranntees less value estimation error and larger performance improvement over the reference policy. We evaluate Uni-RL on a range of standard RL benchmarks across online, offline, and offline-to-online settings. In online RL, Uni-RL achieves higher sample efficiency than both off-policy methods without trust-region updates and on-policy methods with trust-region updates. In offline RL, Uni-RL retains the benefits of in-sample learning while outperforming IVR through better policy extraction. In offline-to-online RL, Uni-RL beats both constraint-based methods and unconstrained approaches by effectively balancing stability and adaptability.

The rapid advancement of Large Language Models (LLMs) has catalyzed the development of multi-agent systems, where multiple LLM-based agents collaborate to solve complex tasks. However, existing systems predominantly rely on centralized coordination, which introduces scalability bottlenecks, limits adaptability, and creates single points of failure. Additionally, concerns over privacy and proprietary knowledge sharing hinder cross-organizational collaboration, leading to siloed expertise. To address these challenges, we propose AgentNet, a decentralized, Retrieval-Augmented Generation (RAG)-based framework that enables LLM-based agents to autonomously evolve their capabilities and collaborate efficiently in a Directed Acyclic Graph (DAG)-structured network. Unlike traditional multi-agent systems that depend on static role assignments or centralized control, AgentNet allows agents to specialize dynamically, adjust their connectivity, and route tasks without relying on predefined workflows. AgentNet’s core design is built upon several key innovations: (1) Fully Decentralized Paradigm: Removing the central orchestrator, allowing agents to coordinate and specialize autonomously, fostering fault tolerance and emergent collective intelligence. (2) Dynamically Evolving Graph Topology: Real-time adaptation of agent connections based on task demands, ensuring scalability and resilience. (3) Adaptive Learning for Expertise Refinement: A retrieval-based memory system that enables agents to continuously update and refine their specialized skills. By eliminating centralized control, AgentNet enhances fault tolerance, promotes scalable specialization, and enables privacy-preserving collaboration across organizations. Through decentralized coordination and minimal data exchange, agents can leverage diverse knowledge sources while safeguarding sensitive information. Experimental results demonstrate that AgentNet outperforms traditional centralized multi-agent systems, significantly improving efficiency, adaptability, and scalability in dynamic environments, making it a promising foundation for next-generation autonomous, privacy-respecting multi-agent ecosystems.

Realignment becomes necessary when a language model (LM) fails to meet expected performance. We propose a flexible realignment framework that supports quantitative control of alignment degree during training and inference. This framework incorporates **Training-time Realignment (TrRa)**, which efficiently realigns the reference model by leveraging the controllable fusion of logits from both the reference and already aligned models. For example, TrRa reduces token usage by **54.63%** on DeepSeek-R1-Distill-Qwen-1.5B without any performance degradation, outperforming DeepScaleR-1.5B’s **33.86%**. To complement TrRa during inference, we introduce a layer adapter that enables **smooth Inference-time Realignment (InRa)**. This adapter is initialized to perform an identity transformation at the bottom layer and is inserted preceding the original layers. During inference, input embeddings are simultaneously processed by the adapter and the original layer, followed by the remaining layers, and then controllably interpolated at the logit level. We upgraded DeepSeek-R1-Distill-Qwen-7B from a slow-thinking model to one that supports both fast and slow thinking, allowing flexible alignment control even **during inference**. By encouraging deeper reasoning, it even surpassed its original performance.

Information retrieval (IR) systems are essential in modern society, aiding users to efficiently locate relevant information through query expansion, document retrieval, ranking, and re-ranking. User feedback from ranked outputs forms a dynamic interaction loop with IR systems, which can be modeled as either one-time or sequential decision-making problems. Over the past decade, deep reinforcement learning (DRL) has emerged as a promising approach to decision-making, leveraging the high model capacity of deep learning for complex tasks. While significant research has explored the application of DRL to IR tasks, several fundamental challenges remain underexplored, including the underlying information theory in DRL settings, the limitations of reinforcement learning methods for industrial IR applications, and the simulation of DRL-based IR systems. Concurrently, the advent of large language models (LLMs) has introduced new opportunities for optimizing and simulating IR systems. Building on the success of the Agent-based IR Workshop at SIGIR 2024, we propose hosting the second Agent-based IR Workshop at SIGIR 2025. This workshop will continue to provide a platform for researchers and practitioners from academia and industry to present cutting-edge advances in DRL-based and LLM-based IR systems from an agent-based perspective. By building on the foundation laid in the first workshop, the 2025 edition aims to delve deeper into emerging research challenges, foster collaborations, and explore innovative applications. Through engaging discussions and insightful presentations, the workshop seeks to further expand the boundaries of IR research and solidify its role as a premier venue for advancing agent-based IR systems.

Interactive recommender systems (IRSs) aim to meet user needs through natural language dialogues, optimizing recommendations with minimal interactions. Typically, IRSs are based on large language models (LLMs). Existing methods generally consist of two stages: decision-making (deciding whether to recommend or ask clarification questions) and action execution (generating recommendations or clarification questions). These methods usually follow a decision-first paradigm, where the model first decides on the action based on past conversations, and then executes the corresponding action. Since LLMs struggle to process a large number of candidate items, the recommendation process is often carried out in collaboration with external recommendation tools, which provide a small candidate set for LLMs to refine.

The past few years have witnessed a growing interest in LLM-based recommender systems (RSs), although their industrial deployment remains in a preliminary stage. Most existing deployments leverage LLMs offline as feature enhancers, generating augmented knowledge for downstream tasks. However, in recommendation scenarios with numerous users and items, even offline knowledge generation with LLMs demands significant time and computational resources. This inefficiency arises from the autoregressive nature of LLMs. A promising solution is speculative decoding, a Draft-Then-Verify approach that increases the number of tokens generated per decoding step. In this work, we first identify recommendation knowledge generation as a highly fitting use case for retrieval-based speculative decoding. Then, we discern its two characteristics: (1) the vast number of items and users in RSs leads to retrieval inefficiency, and (2) RSs exhibit high diversity tolerance for LLM-generated text. Building on these insights, we introduce Lossless Acceleration via Speculative Decoding for LLM-based Recommender Systems (LASER), which features a Customized Retrieval Pool to enhance retrieval efficiency and Relaxed Verification to improve the acceptance rate of draft tokens. LASER achieves a 3-5x speedup on public datasets and saves about 67% of computational resources during the online A/B test on a large-scale advertising scenario with lossless downstream recommendation performance. Our code is available at https://github.com/YunjiaXi/LASER

Recommender systems (RS) are pivotal in managing information overload in modern digital services. A key challenge in RS is efficiently processing vast item pools to deliver highly personalized recommendations under strict latency constraints. Multi-stage cascade ranking addresses this by employing computationally efficient retrieval methods to cover diverse user interests, followed by more precise ranking models to refine the results. In the retrieval stage, multi-channel retrieval is often used to generate distinct item subsets from different candidate generators, leveraging the complementary strengths of these methods to maximize coverage. However, forwarding all retrieved items overwhelms downstream rankers, necessitating truncation. Despite advancements in individual retrieval methods, multi-channel fusion, the process of efficiently merging multi-channel retrieval results, remains underexplored. We are the first to identify and systematically investigate multi-channel fusion in the retrieval stage. Current industry practices often rely on heuristic approaches and manual designs, which often lead to suboptimal performance. Moreover, traditional gradient-based methods like SGD are unsuitable for this task due to the non-differentiable nature of the selection process. In this paper, we explore advanced channel fusion strategies by assigning systematically optimized weights to each channel. We utilize black-box optimization techniques, including the Cross Entropy Method and Bayesian Optimization for global weight optimization, alongside policy gradient-based approaches for personalized merging. Our methods enhance both personalization and flexibility, achieving significant performance improvements across multiple datasets and yielding substantial gains in real-world deployments, offering a scalable solution for optimizing multi-channel fusion in retrieval.

A vast amount of user behavior data is constantly accumulating on today's large recommendation platforms, recording users' various interests and tastes. Preserving knowledge from the old data while new data continually arrives is a vital problem for recommender systems. Existing approaches generally seek to save the knowledge implicitly in the model parameters. However, such a parameter-centric approach lacks scalability and flexibility---the capacity is hard to scale, and the knowledge is inflexible to utilize. Hence, in this work, we propose a framework that turns massive user behavior data to retrievable knowledge (D2K). It is a data-centric approach that is model-agnostic and easy to scale up. Different from only storing unary knowledge such as the user-side or item-side information, D2K propose to store ternary knowledge for recommendation, which is determined by the complete recommendation factors---user, item, and context. The knowledge retrieved by target samples can be directly used to enhance the performance of any recommendation algorithms. Specifically, we introduce a Transformer-based knowledge encoder to transform the old data into knowledge with the user-item-context cross features. A personalized knowledge adaptation unit is devised to effectively exploit the information from the knowledge base by adapting the retrieved knowledge to the target samples. Extensive experiments demonstrate the superiority of the proposed method.

With the rapid development of online services and web applications, recommender systems (RS) have become increasingly indispensable for mitigating information overload and matching users’ information needs by providing personalized suggestions over items. Although the RS research community has made remarkable progress over the past decades, conventional recommendation models (CRM) still have some limitations, e.g., lacking open-domain world knowledge, and difficulties in comprehending users’ underlying preferences and motivations. Meanwhile, large language models (LLM) have shown impressive general intelligence and human-like capabilities for various natural language processing (NLP) tasks, which mainly stem from their extensive open-world knowledge, logical and commonsense reasoning abilities, as well as their comprehension of human culture and society. Consequently, the emergence of LLM is inspiring the design of RS and pointing out a promising research direction, i.e., whether we can incorporate LLM and benefit from their common knowledge and capabilities to compensate for the limitations of CRM. In this article, we conduct a comprehensive survey on this research direction, and draw a bird’s-eye view from the perspective of the whole pipeline in real-world RS. Specifically, we summarize existing research works from two orthogonal aspects: where and how to adapt LLM to RS. For the “ WHERE ” question, we discuss the roles that LLM could play in different stages of the recommendation pipeline, i.e., feature engineering, feature encoder, scoring/ranking function, user interaction, and pipeline controller. For the “ HOW ” question, we investigate the training and inference strategies, resulting in two fine-grained taxonomy criteria, i.e., whether to tune LLM or not during training, and whether to involve CRM for inference. Detailed analysis and general development paths are provided for both “WHERE” and “HOW” questions, respectively. Then, we highlight the key challenges in adapting LLM to RS from three aspects, i.e., efficiency, effectiveness, and ethics. Finally, we summarize the survey and discuss the future prospects.

With the development of multimedia systems, multimodal recommendations are playing an essential role, as they can leverage rich contexts beyond interactions. Existing methods mainly regard multimodal information as an auxiliary, using them to help learn ID features; However, there exist semantic gaps among multimodal content features and ID-based features, for which directly using multimodal information as an auxiliary would lead to misalignment in representations of users and items. In this paper, we first systematically investigate the misalignment issue in multimodal recommendations, and propose a solution named AlignRec. In AlignRec, the recommendation objective is decomposed into three alignments, namely alignment within contents, alignment between content and categorical ID, and alignment between users and items. Each alignment is characterized by a specific objective function and is integrated into our multimodal recommendation framework. To effectively train AlignRec, we propose starting from pre-training the first alignment to obtain unified multimodal features and subsequently training the following two alignments together with these features as input. As it is essential to analyze whether each multimodal feature helps in training and accelerate the iteration cycle of recommendation models, we design three new classes of metrics to evaluate intermediate performance. Our extensive experiments on three real-world datasets consistently verify the superiority of AlignRec compared to nine baselines. We also find that the multimodal features generated by AlignRec are better than currently used ones, which are to be open-sourced in our repository https://github.com/sjtulyf123/AlignRec_CIKM24.

Large language models have been flourishing in the natural language processing (NLP) domain, and their potential for recommendation has been paid much attention to. Despite the intelligence shown by the recommendation-oriented finetuned models, LLMs struggle to fully understand the user behavior patterns due to their innate weakness in interpreting numerical features and the overhead for long context, where the temporal relations among user behaviors, subtle quantitative signals among different ratings, and various side features of items are not well explored. Existing works only fine-tune a sole LLM on given text data without introducing that important information to it, leaving these problems unsolved. In this paper, we propose ELCoRec to Enhance Language understanding with Co-Propagation of numerical and categorical features for Recommendation. Concretely, we propose to inject the preference understanding capability into LLM via a GAT expert model where the user preference is better encoded by parallelly propagating the temporal relations, and rating signals as well as various side information of historical items. The parallel propagation mechanism could stabilize heterogeneous features and offer an informative user preference encoding, which is then injected into the language models via soft prompting at the cost of a single token embedding. To further obtain the user's recent interests, we proposed a novel Recent interaction Augmented Prompt (RAP) template. Experiment results over three datasets against strong baselines validate the effectiveness of ELCoRec.

Knowledge Tracing (KT) aims to determine whether students will respond correctly to the next question, which is a crucial task in intelligent tutoring systems (ITS). In educational KT scenarios, transductive ID-based methods often face severe data sparsity and cold start problems, where interactions between individual students and questions are sparse, and new questions and concepts consistently arrive in the database. In addition, existing KT models only implicitly consider the correlation between concepts and questions, lacking direct modeling of the more complex relationships in the heterogeneous graph of concepts and questions. In this paper, we propose a Structure-aware INductive Knowledge Tracing model with large language model (dubbed SINKT), which, for the first time, introduces large language models (LLMs) and realizes inductive knowledge tracing. Firstly, SINKT utilizes LLMs to introduce structural relationships between concepts and constructs a hetero- geneous graph for concepts and questions. Secondly, by encoding concepts and questions with LLMs, SINKT incorporates semantic information to aid prediction. Finally, SINKT predicts the student's response to the target question by interacting with the student's knowledge state and the question representation. Experiments on four real-world datasets demonstrate that SINKT achieves state-of-the-art performance among 12 existing transductive KT models. Additionally, we explore the performance of SINKT on the inductive KT task and provide insights into various modules.

Conversational recommender systems (CRSs) aim to capture user preferences and provide personalized recommendations through multi-round natural language dialogues. However, most existing CRS models mainly focus on dialogue comprehension and preferences mining from the current dialogue session, overlooking user preferences in historical dialogue sessions. The preferences embedded in historical sessions and the current session exhibit continuity and sequentiality, and we refer to such CRSs as sequential CRSs. In this work, we leverage memory-enhanced LLMs to model the preference continuity, addressing two key issues: (1) redundancy and noise in historical dialogue sessions, and (2) the cold-start users problem. Thus, we propose a Memory-enhanced Conversational Recommender System Framework with Large Language Models (dubbed MemoCRS), consisting of user-specific memory and general memory. User-specific memory is tailored to each user's interests and uses an entity-based memory bank to refine preferences and retrieve relevant memory, thereby reducing the redundancy and noise of historical sessions. The general memory, encapsulating collaborative knowledge and reasoning guidelines, can provide shared knowledge for users, especially cold-start users. With the above memory, LLMs are empowered to deliver more precise and tailored recommendations for each user. Extensive experiments on Chinese and English datasets demonstrate MemoCRS's effectiveness.

Crafting effective features is a crucial yet labor-intensive and domain-specific task within machine learning pipelines. Fortunately, recent advancements in Large Language Models (LLMs) have shown promise in automating various data science tasks, including feature engineering. But despite this potential, evaluations thus far are primarily based on the end performance of a complete ML pipeline, providing limited insight into precisely how LLMs behave relative to human experts in feature engineering. To address this gap, we propose ELF-Gym, a framework for Evaluating LLM-generated Features. We curated a new dataset from historical Kaggle competitions, including 251 golden features used by top-performing teams. ELF-Gym then quantitatively evaluates LLM-generated features by measuring their impact on downstream model performance as well as their alignment with expert-crafted features through semantic and functional similarity assessments. This approach provides a more comprehensive evaluation of disparities between LLMs and human experts, while offering valuable insights into specific areas where LLMs may have room for improvement. For example, using ELF-Gym we empirically demonstrate that, in the best-case scenario, LLMs can semantically capture approximately 56% of the golden features, but at the more demanding implementation level this overlap drops to 13%. Moreover, in other cases LLMs may fail completely, particularly on datasets that require complex features, indicating broad potential pathways for improvement.

Accurate prediction of multi-agent spatiotemporal systems is critical to various real-world applications, such as autonomous driving, sports, and multiplayer games.Unfortunately, modeling multiagent trajectories is challenging due to its complicated, interactive, and multi-modal nature.Recently, diffusion models have achieved great success in modeling multi-modal distribution and trajectory generation, showing promising ability in resolving this problem.Motivated by this, in this paper, we propose a novel multi-agent trajectory prediction framework, dubbed Scalable Diffusion Transformer (SDT), which is naturally designed to learn the complicated distribution and implicit interactions among agents.We evaluate SDT on a set of real-world benchmark datasets and compare it with representative baseline methods, which demonstrates the better multi-agent trajectory prediction ability of SDT in terms of accuracy and diversity.

Many real-world problems can be formulated as a constrained Traveling Salesman Problem (TSP).However, the constraints are always complex and numerous, making the TSPs challenging to solve.When the number of complicated constraints grows, it is timeconsuming for traditional heuristic algorithms to avoid illegitimate outcomes.Learning-based methods provide an alternative to solve TSPs in a soft manner, which also supports GPU acceleration to generate solutions quickly.Nevertheless, the soft manner inevitably results in difficulty solving hard-constrained problems with learning algorithms, and the conflicts between legality and optimality may substantially affect the optimality of the solution.To overcome this problem and to have an effective solution against hard constraints, we proposed a novel learning-based method, MUSLA, that uses multi-step looking-ahead information as the feature to improve the legality of TSP with Time Windows (TSPTW) solutions.Besides, we constructed TSPTW datasets with hard constraints in order to accurately evaluate and benchmark the statistical performance of various approaches, which can serve the community for future research.With comprehensive experiments on diverse datasets, MUSLA outperforms existing baselines and shows generalizability potential.

Recent progress in vision language foundation models has shown their ability to understand multimodal data and resolve complicated vision language tasks, including robotics manipulation. We seek a straightforward way of making use of existing vision-language models (VLMs) with simple fine-tuning on robotics data. To this end, we derive a simple and novel vision-language manipulation framework, dubbed RoboFlamingo, built upon the open-source VLMs, OpenFlamingo. Unlike prior works, RoboFlamingo utilizes pre-trained VLMs for single-step vision-language comprehension, models sequential history information with an explicit policy head, and is slightly fine-tuned by imitation learning only on language-conditioned manipulation datasets. Such a decomposition provides RoboFlamingo the flexibility for open-loop control and deployment on low-performance platforms. By exceeding the state-of-the-art performance with a large margin on the tested benchmark, we show RoboFlamingo can be an effective and competitive alternative to adapt VLMs to robot control. Our extensive experimental results also reveal several interesting conclusions regarding the behavior of different pre-trained VLMs on manipulation tasks. We believe RoboFlamingo has the potential to be a cost-effective and easy-to-use solution for robotics manipulation, empowering everyone with the ability to fine-tune their own robotics policy. Our code will be made public upon acceptance.

In this study, we investigate the DIstribution Correction Estimation (DICE) methods, an important line of work in offline reinforcement learning (RL) and imitation learning (IL). DICE-based methods impose state-action-level behavior constraint, which is an ideal choice for offline learning. However, they typically perform much worse than current state-of-the-art (SOTA) methods that solely use action-level behavior constraint. After revisiting DICE-based methods, we find there exist two gradient terms when learning the value function using true-gradient update: forward gradient (taken on the current state) and backward gradient (taken on the next state). Using forward gradient bears a large similarity to many offline RL methods, and thus can be regarded as applying action-level constraint. However, directly adding the backward gradient may degenerate or cancel out its effect if these two gradients have conflicting directions. To resolve this issue, we propose a simple yet effective modification that projects the backward gradient onto the normal plane of the forward gradient, resulting in an orthogonal-gradient update, a new learning rule for DICE-based methods. We conduct thorough theoretical analyses and find that the projected backward gradient brings state-level behavior regularization, which reveals the mystery of DICE-based methods: the value learning objective does try to impose state-action-level constraint, but needs to be used in a corrected way. Through toy examples and extensive experiments on complex offline RL and IL tasks, we demonstrate that DICE-based methods using orthogonal-gradient updates achieve SOTA performance and great robustness.

In computer animation, driving a simulated character with lifelike motion is challenging. Current generative models, though able to generalize to diverse motions, often pose challenges to the responsiveness of end-user control. To address these issues, we introduce RACon: Retrieval-Augmented Simulated Character Locomotion Control. Our end-to-end hierarchical reinforcement learning method utilizes a retriever and a motion controller. The retriever searches motion experts from a user-specified database in a task-oriented fashion, which boosts the responsiveness to the user’s control. The selected motion experts and the manipulation signal are then transferred to the controller to drive the simulated character. In addition, a retrieval-augmented discriminator is designed to stabilize the training process. Our method surpasses existing techniques in both quality and quantity in locomotion control, as demonstrated in our empirical study. Moreover, by switching extensive databases for retrieval, it can adapt to distinctive motion types at run time. We will release our code upon acceptance.

In offline reinforcement learning (RL), the performance of the learned policy highly depends on the quality of offline datasets. However, the offline dataset contains very limited optimal trajectories in many cases. This poses a challenge for offline RL algorithms, as agents must acquire the ability to transit to high-reward regions. To address this issue, we introduce Diffusionbased Trajectory Stitching (DiffStitch), a novel diffusion-based data augmentation pipeline that systematically generates stitching transitions between trajectories. DiffStitch effectively connects low-reward trajectories with high-reward trajectories, forming globally optimal trajectories and thereby mitigating the challenges faced by offline RL algorithms in learning trajectory stitching. Empirical experiments conducted on D4RL datasets demonstrate the effectiveness of our pipeline across RL methodologies. Notably, DiffStitch demonstrates substantial enhancements in the performance of one-step methods(IQL), imitation learning methods(TD3+BC) and trajectory optimization methods(DT). Our code is publicly available at https://github.com/guangheli12/DiffStitch

Recent works like Tree-of-Thought (ToT) and Reasoning via Planning (RAP) aim to augment the multi-step reasoning capabilities of LLMs by using tree-search algorithms. These methods rely on prompting a pre-trained model to serve as a value function and focus on problems with low search depth. As a result, these methods cannot benefit from in-domain training and only rely on pretraining process — they will not work in domains where the pre-trained LLM does not have enough knowledge to serve as an effective value function or in domains that require long-horizon planning. To address these limitations, we present an AlphaZero-like tree-search learning framework for LLMs (termed TS-LLM), systematically illustrating how tree-search with a learned value function can guide LLM decoding. TS-LLM distinguishes itself in two key ways. (1) Leveraging a learned value function and AlphaZero-like algorithms, our approach can be generally adaptable to a wide range of tasks, language models of any size, and tasks of varying search depths. (2) Our approach can guide LLMs during both inference and training, iteratively improving the LLMs. Empirical results across reasoning, planning, alignment, and decision-making tasks show that TS-LLM outperforms existing approaches and can handle trees with a depth of 64.

Recommender systems play important roles in various applications such as e-commerce, social media, etc. Conventional recommendation methods usually model the collaborative signals within the tabular representation space. Despite the personalization modeling and the efficiency, the latent semantic dependencies are omitted. Methods that introduce semantics into recommendation then emerge, injecting knowledge from the semantic representation space where the general language understanding are compressed. However, existing semantic-enhanced recommendation methods focus on aligning the two spaces, during which the representations of the two spaces tend to get close while the unique patterns are discarded and not well explored. In this paper, we propose DisCo to Disentangle the unique patterns from the two representation spaces and Collaborate the two spaces for recommendation enhancement, where both the specificity and the consistency of the two spaces are captured. Concretely, we propose 1) a dual-side attentive network to capture the intra-domain patterns and the inter-domain patterns, 2) a sufficiency constraint to preserve the task-relevant information of each representation space and filter out the noise, and 3) a disentanglement constraint to avoid the model from discarding the unique information. These modules strike a balance between disentanglement and collaboration of the two representation spaces to produce informative pattern vectors, which could serve as extra features and be appended to arbitrary recommendation backbones for enhancement. Experiment results validate the superiority of our method against different models and the compatibility of DisCo over different backbones. Various ablation studies and efficiency analysis are also conducted to justify each model component.

Diffusion models have garnered widespread attention in Reinforcement Learning (RL) for their powerful expressiveness and multimodality. It has been verified that utilizing diffusion policies can significantly improve the performance of RL algorithms in continuous control tasks by overcoming the limitations of unimodal policies, such as Gaussian policies. Furthermore, the multimodality of diffusion policies also shows the potential of providing the agent with enhanced exploration capabilities. However, existing works mainly focus on applying diffusion policies in offline RL, while their incorporation into online RL has been less investigated. The diffusion model's training objective, known as the variational lower bound, cannot be applied directly in online RL due to the unavailability of 'good' samples (actions). To harmonize the diffusion model with online RL, we propose a novel model-free diffusion-based online RL algorithm named Q-weighted Variational Policy Optimization (QVPO). Specifically, we introduce the Q-weighted variational loss and its approximate implementation in practice. Notably, this loss is shown to be a tight lower bound of the policy objective. To further enhance the exploration capability of the diffusion policy, we design a special entropy regularization term. Unlike Gaussian policies, the log-likelihood in diffusion policies is inaccessible; thus this entropy term is nontrivial. Moreover, to reduce the large variance of diffusion policies, we also develop an efficient behavior policy through action selection. This can further improve its sample efficiency during online interaction. Consequently, the QVPO algorithm leverages the exploration capabilities and multimodality of diffusion policies, preventing the RL agent from converging to a sub-optimal policy. To verify the effectiveness of QVPO, we conduct comprehensive experiments on MuJoCo continuous control benchmarks. The final results demonstrate that QVPO achieves state-of-the-art performance in terms of both cumulative reward and sample efficiency.

Learning a generalist embodied agent capable of completing multiple tasks poses challenges, primarily stemming from the scarcity of action-labeled robotic datasets. In contrast, a vast amount of human videos exist, capturing intricate tasks and interactions with the physical world. Promising prospects arise for utilizing actionless human videos for pre-training and transferring the knowledge to facilitate robot policy learning through limited robot demonstrations. However, it remains a challenge due to the domain gap between humans and robots. Moreover, it is difficult to extract useful information representing the dynamic world from human videos, because of its noisy and multimodal data structure. In this paper, we introduce a novel framework to tackle these challenges, which leverages a unified discrete diffusion to combine generative pre-training on human videos and policy fine-tuning on a small number of action-labeled robot videos. We start by compressing both human and robot videos into unified video tokens. In the pre-training stage, we employ a discrete diffusion model with a mask-and-replace diffusion strategy to predict future video tokens in the latent space. In the fine-tuning stage, we harness the imagined future videos to guide low-level action learning with a limited set of robot data. Experiments demonstrate that our method generates high-fidelity future videos for planning and enhances the fine-tuned policies compared to previous state-of-the-art approaches with superior performance.

One important property of DIstribution Correction Estimation (DICE) methods is that the solution is the optimal stationary distribution ratio between the optimized and data collection policy. In this work, we show that DICE-based methods can be viewed as a transformation from the behavior distribution to the optimal policy distribution. Based on this, we propose a novel approach, Diffusion-DICE, that directly performs this transformation using diffusion models. We find that the optimal policy's score function can be decomposed into two terms: the behavior policy's score function and the gradient of a guidance term which depends on the optimal distribution ratio. The first term can be obtained from a diffusion model trained on the dataset and we propose an in-sample learning objective to learn the second term. Due to the multi-modality contained in the optimal policy distribution, the transformation in Diffusion-DICE may guide towards those local-optimal modes. We thus generate a few candidate actions and carefully select from them to achieve global-optimum. Different from all other diffusion-based offline RL methods, the \textit{guide-then-select} paradigm in Diffusion-DICE only uses in-sample actions for training and brings minimal error exploitation in the value function. We use a didatic toycase example to show how previous diffusion-based methods fail to generate optimal actions due to leveraging these errors and how Diffusion-DICE successfully avoid that. We then conduct extensive experiments on benchmark datasets to show the strong performance of Diffusion-DICE.

Given a relational database (RDB), how can we predict missing column values in some target table of interest? Although RDBs store vast amounts of rich, informative data spread across interconnected tables, the progress of predictive machine learning models as applied to such tasks arguably falls well behind advances in other domains such as computer vision or natural language processing. This deficit stems, at least in part, from the lack of established/public RDB benchmarks as needed for training and evaluation purposes. As a result, related model development thus far often defaults to tabular approaches trained on ubiquitous single-table benchmarks, or on the relational side, graph-based alternatives such as GNNs applied to a completely different set of graph datasets devoid of tabular characteristics. To more precisely target RDBs lying at the nexus of these two complementary regimes, we explore a broad class of baseline models predicated on: (i) converting multi-table datasets into graphs using various strategies equipped with efficient subsampling, while preserving tabular characteristics; and (ii) trainable models with well-matched inductive biases that output predictions based on these input subgraphs. Then, to address the dearth of suitable public benchmarks and reduce siloed comparisons, we assemble a diverse collection of (i) large-scale RDB datasets and (ii) coincident predictive tasks. From a delivery standpoint, we operationalize the above four dimensions (4D) of exploration within a unified, scalable open-source toolbox called 4DBInfer; please see https://github.com/awslabs/multi-table-benchmark .

Zero-shot coordination (ZSC) is a new cooperative multi-agent reinforcement learning (MARL) challenge that aims to train an ego agent to work with diverse, unseen partners during deployment. The significant difference between the deployment-time partners' distribution and the training partners' distribution determined by the training algorithm makes ZSC a unique out-of-distribution (OOD) generalization challenge. The potential distribution gap between evaluation and deployment-time partners leads to inadequate evaluation, which is exacerbated by the lack of appropriate evaluation metrics. In this paper, we present ZSC-Eval, the first evaluation toolkit and benchmark for ZSC algorithms. ZSC-Eval consists of: 1) Generation of evaluation partner candidates through behavior-preferring rewards to approximate deployment-time partners' distribution; 2) Selection of evaluation partners by Best-Response Diversity (BR-Div); 3) Measurement of generalization performance with various evaluation partners via the Best-Response Proximity (BR-Prox) metric. We use ZSC-Eval to benchmark ZSC algorithms in Overcooked and Google Research Football environments and get novel empirical findings. We also conduct a human experiment of current ZSC algorithms to verify the ZSC-Eval's consistency with human evaluation. ZSC-Eval is now available at https://github.com/sjtu-marl/ZSC-Eval.

Language models as intelligent agents push the boundaries of sequential decision-making agents but struggle with limited knowledge of environmental dynamics and exponentially huge action space. Recent efforts like GLAM and TWOSOME manually constrain the action space to a restricted subset and employ reinforcement learning to align agents' knowledge with specific environments. However, they overlook fine-grained credit assignments for intra-action tokens, which is essential for efficient language agent optimization, and rely on human's prior knowledge to restrict action space. This paper proposes decomposing language agent optimization from the action level to the token level, offering finer supervision for each intra-action token and manageable optimization complexity in environments with unrestricted action spaces. Beginning with the simplification of flattening all actions, we theoretically explore the discrepancies between action-level optimization and this naive token-level optimization. We then derive the Bellman backup with Action Decomposition (BAD) to integrate credit assignments for both intra-action and inter-action tokens, effectively eliminating the discrepancies. Implementing BAD within the PPO algorithm, we introduce Policy Optimization with Action Decomposition (POAD). POAD benefits from a finer-grained credit assignment process and lower optimization complexity, leading to enhanced learning efficiency and generalization abilities in aligning language agents with interactive environments. We validate POAD across diverse testbeds, with results affirming the advantages of our approach and the correctness of our theoretical analysis. The source code can be accessed directly with this link: https://github.com/morning9393/ADRL.

Offline reinforcement learning (RL) aims to learn policies from pre-existing datasets without further interactions, making it a challenging task. Q-learning algorithms struggle with extrapolation errors in offline settings, while supervised learning methods are constrained by model expressiveness. Recently, diffusion models (DMs) have shown promise in overcoming these limitations in single-agent learning, but their application in multi-agent scenarios remains unclear. Generating trajectories for each agent with independent DMs may impede coordination, while concatenating all agents’ information can lead to low sample efficiency. Accordingly, we propose MADiff, which is realized with an attention-based diffusion model to model the complex coordination among behaviors of multiple agents. To our knowledge, MADiff is the first diffusion-based multi-agent learning framework, functioning as both a decentralized policy and a centralized controller. During decentralized executions, MADiff simultaneously performs teammate modeling, and the centralized controller can also be applied in multi-agent trajectory predictions. Our experiments demonstrate that MADiff outperforms baseline algorithms across various multi-agent learning tasks, highlighting its effectiveness in modeling complex multi-agent interactions.

Click-through rate (CTR) prediction plays as a core function module in various personalized online services. The traditional ID-based models for CTR prediction take as inputs the one-hot encoded ID features of tabular modality, which capture the collaborative signals via feature interaction modeling. But the one-hot encoding discards the semantic information included in the textual features. Recently, the emergence of Pretrained Language Models (PLMs) has given rise to another paradigm, which takes as inputs the sentences of textual modality obtained by hard prompt templates and adopts PLMs to extract the semantic knowledge. However, PLMs often face challenges in capturing field-wise collaborative signals and distinguishing features with subtle textual differences. In this paper, to leverage the benefits of both paradigms and meanwhile overcome their limitations, we propose to conduct Fine-grained feature-level ALignment between ID-based Models and Pretrained Language Models (FLIP) for CTR prediction. Unlike most methods that solely rely on global views through instance-level contrastive learning, we design a novel jointly masked tabular/language modeling task to learn fine-grained alignment between tabular IDs and word tokens. Specifically, the masked data of one modality (i.e., IDs and tokens) has to be recovered with the help of the other modality, which establishes the feature-level interaction and alignment via sufficient mutual information extraction between dual modalities. Moreover, we propose to jointly finetune the ID-based model and PLM by adaptively combining the output of both models, thus achieving superior performance in downstream CTR prediction tasks. Extensive experiments on three real-world datasets demonstrate that FLIP outperforms SOTA baselines, and is highly compatible with various ID-based models and PLMs. The code is available12.

Recommender system plays a vital role in various online services. However, its insulated nature of training and deploying separately within a specific closed domain limits its access to open-world knowledge. Recently, the emergence of large language models (LLMs) has shown promise in bridging this gap by encoding extensive world knowledge and demonstrating reasoning capabilities. Nevertheless, previous attempts to directly use LLMs as recommenders cannot meet the inference latency demand of industrial recommender systems. In this work, we propose an Open-World Knowledge Augmented Recommendation Framework with Large Language Models, dubbed KAR, to acquire two types of external knowledge from LLMs — the reasoning knowledge on user preferences and the factual knowledge on items. We introduce factorization prompting to elicit accurate reasoning on user preferences. The generated reasoning and factual knowledge are effectively transformed and condensed into augmented vectors by a hybrid-expert adaptor in order to be compatible with the recommendation task. The obtained vectors can then be directly used to enhance the performance of any recommendation model. We also ensure efficient inference by preprocessing and prestoring the knowledge from the LLM. Extensive experiments show that KAR significantly outperforms the state-of-the-art baselines and is compatible with a wide range of recommendation algorithms. We deploy KAR to Huawei’s news and music recommendation platforms and gain a 7% and 1.7% improvement in the online A/B test, respectively.

Information retrieval (IR) systems have become an essential component in modern society to help users find useful information, which consists of a series of processes including query expansion, item recall, item ranking and re-ranking, etc. Based on the ranked information list, users can provide their feedbacks. Such an interaction process between users and IR systems can be naturally formulated as a decision-making problem, which can be either one-step or sequential. In the last ten years, deep reinforcement learning (DRL) has become a promising direction for decision-making, since DRL utilizes the high model capacity of deep learning for complex decision-making tasks. On the one hand, there have been emerging research works focusing on leveraging DRL for IR tasks. However, the fundamental information theory under DRL settings, the challenge of RL methods for Industrial IR tasks, or the simulations of DRL-based IR systems, has not been deeply investigated. On the other hand, the emerging LLM provides new opportunities for optimizing and simulating IR systems. To this end, we propose the first Agent-based IR workshop at SIGIR 2024, as a continuation from one of the most successful IR workshops, DRL4IR. It provides a venue for both academia researchers and industry practitioners to present the recent advances of both DRL-based IR systems and LLM-based IR systems from the agent-based IR's perspective, to foster novel research, interesting findings, and new applications.

Several large language model (LLM) agents have been constructed for diverse purposes such as web navigation and online shopping, leveraging the broad knowledge and text comprehension capabilities of LLMs. Many of these works rely on in-context examples to achieve generalization without requiring fine-tuning. However, few have addressed the challenge of selecting and effectively utilizing these examples. Recent approaches have introduced trajectory-level retrieval with task meta-data and the use of trajectories as in-context examples to enhance overall performance in some sequential decision making tasks like computer control. Nevertheless, these methods face issues like plausible examples retrieved without task-specific state transition dynamics and long input with plenty of irrelevant context due to using complete trajectories. In this paper, we propose a novel framework (TRAD) to tackle these problems. TRAD first employs Thought Retrieval for step-level demonstration selection through thought matching, enhancing the quality of demonstrations and reducing irrelevant input noise. Then, Aligned Decision is introduced to complement retrieved demonstration steps with their preceding or subsequent steps, providing tolerance for imperfect thought and offering a balance between more context and less noise. Extensive experiments on ALFWorld and Mind2Web benchmarks demonstrate that TRAD not only surpasses state-of-the-art models but also effectively reduces noise and promotes generalization. Furthermore, TRAD has been deployed in real-world scenarios of a global business insurance company and yields an improved success rate of robotic process automation. Our codes are available at: https://github.com/skyriver-2000/TRAD-Official.

Large language models (LLMs) have achieved great success in general domains of natural language processing. In this paper, we bring LLMs to the realm of geoscience with the objective of advancing research and applications in this field. To this end, we present the first-ever LLM in geoscience, K2, alongside a suite of resources developed to further promote LLM research within geoscience. For instance, we have curated the first geoscience instruction tuning dataset, GeoSignal, which aims to align LLM responses to geoscience-related user queries. Additionally, we have established the first geoscience benchmark, GeoBench, to evaluate LLMs in the context of geoscience. In this work, we experiment with a complete recipe to adapt a pre-trained general-domain LLM to the geoscience domain. Specifically, we further train the LLaMA-7B model on 5.5B tokens of geoscience text corpus, including over 1 million pieces of geoscience literature, and utilize GeoSignal's supervised data to fine-tune the model. Moreover, we share a protocol that can efficiently gather domain-specific data and construct domain-supervised data, even in situations where manpower is scarce. Meanwhile, we equip K2 with the abilities of using tools to be a naive geoscience aide. Experiments conducted on the GeoBench demonstrate the effectiveness of our approach and datasets on geoscience knowledge understanding and utilization.We open-source all the training data and K2 model checkpoints at https://github.com/davendw49/k2

Ranking items regarding individual user interests is a core technique of multiple downstream tasks such as recommender systems. Learning such a personalized ranker typically relies on the implicit feedback from users' past click-through behaviors. However, collected feedback is biased toward previously highly-ranked items and directly learning from it would result in "rich-get-richer" phenomena. In this paper, we propose a simple yet sufficient unbiased learning-to-rank paradigm named InfoRank that aims to simultaneously address both position and popularity biases. We begin by consolidating the impacts of those biases into a single observation factor, thereby providing a unified approach to addressing bias-related issues. Subsequently, we minimize the mutual information between the observation estimation and the relevance estimation conditioned on the input features. By doing so, our relevance estimation can be proved to be free of bias. To implement InfoRank, we first incorporate an attention mechanism to capture latent correlations within user-item features, thereby generating estimations of observation and relevance. We then introduce a regularization term, grounded in conditional mutual information, to promote conditional independence between relevance estimation and observation estimation. Experimental evaluations conducted across three extensive recommendation and search datasets reveal that InfoRank learns more precise and unbiased ranking strategies.

Click-through rate (CTR) prediction has become increasingly indispensable for various Internet applications. Traditional CTR models convert the multi-field categorical data into ID features via one-hot encoding, and extract the collaborative signals among features. Such a paradigm suffers from the problem of semantic information loss. Another line of research explores the potential of pretrained language models (PLMs) for CTR prediction by converting input data into textual sentences through hard prompt templates. Although semantic signals are preserved, they generally fail to capture the collaborative information (e.g., feature interactions, pure ID features), not to mention the unacceptable inference overhead brought by the huge model size. In this paper, we aim to model both the semantic knowledge and collaborative knowledge for accurate CTR estimation, and meanwhile address the inference inefficiency issue. To benefit from both worlds and close their gaps, we propose a novel model-agnostic framework (i.e., ClickPrompt), where we incorporate CTR models to generate interaction-aware soft prompts for PLMs. We design a prompt-augmented masked language modeling (PA-MLM) pretraining task, where PLM has to recover the masked tokens based on the language context, as well as the soft prompts generated by CTR model. The collaborative and semantic knowledge from ID and textual features would be explicitly aligned and interacted via the prompt interface. Then, we can either tune the CTR model with PLM for superior performance, or solely tune the CTR model without PLM for inference efficiency. Experiments on four real-world datasets validate the effectiveness of ClickPrompt compared with existing baselines.

With large language models (LLMs) achieving remarkable breakthroughs in NLP domains, LLM-enhanced recommender systems have received much attention and have been actively explored currently. In this paper, we focus on adapting and empowering a pure large language model for zero-shot and few-shot recommendation tasks. First and foremost, we identify and formulate the lifelong sequential behavior incomprehension problem for LLMs in recommendation domains, i.e., LLMs fail to extract useful information from a textual context of long user behavior sequence, even if the length of context is far from reaching the context limitation of LLMs. To address such an issue and improve the recommendation performance of LLMs, we propose a novel framework, namely Retrieval enhanced Large Language models (ReLLa) for recommendation tasks in both zero-shot and few-shot settings. For zero-shot recommendation, we perform semantic user behavior retrieval (SUBR) to improve the data quality of testing samples, which greatly reduces the difficulty for LLMs to extract the essential knowledge from user behavior sequences. As for few-shot recommendation, we further design retrieval-enhanced instruction tuning (ReiT) by adopting SUBR as a data augmentation technique for training samples. Specifically, we develop a mixed training dataset consisting of both the original data samples and their retrieval-enhanced counterparts. We conduct extensive experiments on three real-world public datasets to demonstrate the superiority of ReLLa compared with existing baseline models, as well as its capability for lifelong sequential behavior comprehension. To be highlighted, with only less than 10% training samples, few-shot ReLLa can outperform traditional CTR models that are trained on the entire training set (e.g., DCNv2, DIN, SIM).

We primarily focus on the field of multi-scenario recommendation, which poses a significant challenge in effectively leveraging data from different scenarios to enhance predictions in scenarios with limited data. Current mainstream efforts mainly center around innovative model network architectures, with the aim of enabling the network to implicitly acquire knowledge from diverse scenarios. However, the uncertainty of implicit learning in networks arises from the absence of explicit modeling, leading to not only difficulty in training but also incomplete user representation and suboptimal performance. Furthermore, through causal graph analysis, we have discovered that the scenario itself directly influences click behavior, yet existing approaches directly incorporate data from other scenarios during the training of the current scenario, leading to prediction biases when they directly utilize click behaviors from other scenarios to train models. To address these problems, we propose the Multi-Scenario Causal-driven Adaptive Network M-scan). This model incorporates a Scenario-Aware Co-Attention mechanism that explicitly extracts user interests from other scenarios that align with the current scenario. Additionally, it employs a Scenario Bias Eliminator module utilizing causal counterfactual inference to mitigate biases introduced by data from other scenarios. Extensive experiments on two public datasets demonstrate the efficacy of our M-scan compared to the existing baseline models.

How to construct imperceptible (realistic) fake samples is critical in adversarial attacks. Due to the sample feature diversity of a recommender system (containing both discrete and continuous features), traditional gradient-based adversarial attack methods may fail to construct realistic fake samples. Meanwhile, most recommendation models adopt click-through rate (CTR) predictors, which usually utilize black-box deep models with discrete features as input. Thus, how to efficiently construct realistic fake samples for black-box recommender systems is still full of challenges. In this article, we propose a hierarchical adversarial attack method against black-box CTR models via generating realistic fake samples, named CTRAttack. To better train the generation network, the weights of its embedding layer are shared with those of the substitute model, with both the similarity loss and classification loss used to update the generation network. To ensure that the discrete features of the generated fake samples are all real, we first adopt the similarity loss to ensure that the distribution of the generated perturbed samples is sufficiently close to the distribution of the real features, and then the nearest neighbor algorithm is used to retrieve the most appropriate features for non-existent discrete features from the candidate instance set. Extensive experiments demonstrate that CTRAttack can not only effectively attack the black-box recommender systems but also improve the robustness of these models while maintaining prediction accuracy.

Real-Time Bidding (RTB) is one of the most important forms of online advertising, where an auction is hosted in real time to sell the individual ad impression. How to design an automated bidding strategy in response to the dynamic auction environment is crucial for improving user experience, protecting the interests of advertisers, and promoting the long-term development of the advertising platform. As an exciting topic in the real-world industry, it has attracted great research interest from several disciplines, most notably data science. There have been abundant studies on bidding strategy design which are based on the large volume of historical ad requests. Despite its popularity and significance, few works provide a summary for bid optimization. In this survey, we present the latest overview of the recent works to shed light on the optimization techniques where most of them are validated in practice. We first explore the optimization problem in different works, explaining how these different settings affect the bidding strategy designs. Then, some forms of bidding functions and specific optimization techniques are illustrated. Further, we specifically discuss a new trend about bidding in first-price auctions, which have gradually become popular in recent years. From this survey, both practitioners and researchers can gain insights of the challenges and future prospects of bid optimization in RTB.

As a critical task for large-scale commercial recommender systems, reranking rearranges items in the initial ranking lists from the previous ranking stage to better meet users’ demands. Foundational work in reranking has shown the potential of improving recommendation results by uncovering mutual influence among items. However, rather than considering the context of initial lists as most existing methods do, an ideal reranking algorithm should consider the counterfactual context— the position and the alignment of the items in the reranked lists . In this work, we propose a novel pairwise reranking framework, Utility-oriented Reranking with Counterfactual Context (URCC), which maximizes the overall utility after reranking efficiently. Specifically, we first design a utility-oriented evaluator, which applies Bi-LSTM and graph attention mechanism to estimate the listwise utility via the counterfactual context modeling. Then, under the guidance of the evaluator, we propose a pairwise reranker model to find the most suitable position for each item by swapping misplaced item pairs. Extensive experiments on two benchmark datasets and a proprietary real-world dataset demonstrate that URCC significantly outperforms the state-of-the-art models in terms of both relevance-based metrics and utility-based metrics.

Recommender systems are expected to be assistants that help human users find relevant information automatically without explicit queries. As recommender systems evolve, increasingly sophisticated learning techniques are applied and have achieved better performance in terms of user engagement metrics such as clicks and browsing time. The increase in the measured performance, however, can have two possible attributions: a better understanding of user preferences, and a more proactive ability to utilize human bounded rationality to seduce user over-consumption. A natural following question is whether current recommendation algorithms are manipulating user preferences. If so, can we measure the manipulation level? In this article, we present a general framework for benchmarking the degree of manipulations of recommendation algorithms, in both slate recommendation and sequential recommendation scenarios. The framework consists of four stages, initial preference calculation, training data collection, algorithm training and interaction, and metrics calculation that involves two proposed metrics, Manipulation Score and Preference Shift. We benchmark some representative recommendation algorithms in both synthetic and real-world datasets under the proposed framework. We have observed that a high online click-through rate does not necessarily mean a better understanding of user initial preference, but ends in prompting users to choose more documents they initially did not favor. Moreover, we find that the training data have notable impacts on the manipulation degrees, and algorithms with more powerful modeling abilities are more sensitive to such impacts. The experiments also verified the usefulness of the proposed metrics for measuring the degree of manipulations. We advocate that future recommendation algorithm studies should be treated as an optimization problem with constrained user preference manipulations.

Extending from sequential recommendation models, in this article, we present a novel framework named Search-based Time-Aware Recommendation (STARec), which first retrieves the historical behaviors of the given user through a search-based retriever and then captures the user’s evolving demands over time through a time-aware sequential network. We notice that the key insight of STARec is to use the feature and labels to augment the representations, and thus the effectiveness of STARec relies on the acquisition of rich browsing records of the target user and powerful representation of each browsed item and thus its performance could heavily drop regarding long-tail users and items. To this end, we extend STARec by constructing a graph upon the user–item interactions and leveraging the graph structure to enhance the representation learning. We call this extended version Search-based Time-Aware Graph-Enhanced Recommendation (STAGE). We conduct extensive experiments on three real-world datasets and STARec achieves consistent superiority. We further compare STAGE against STARec long-tail users and our results demonstrate that STAGE could outperform STARec at most cases. Results of online A/B tests show that STARec and STAGE achieve an average click-through rate improvement of around 6% and 1.5% in the two main item recommendation scenarios, respectively. 1

With the development of the online education system, personalized education recommendation has played an essential role. In this paper, we focus on developing path recommendation systems that aim to generating and recommending an entire learning path to the given user in each session. Noticing that existing approaches fail to consider the correlations of concepts in the path, we propose a novel framework named Set-to-Sequence Ranking-based Concept-aware Learning Path Recommendation (SRC), which formulates the recommendation task under a set-to-sequence paradigm. Specifically, we first design a concept-aware encoder module which can capture the correlations among the input learning concepts. The outputs are then fed into a decoder module that sequentially generates a path through an attention mechanism that handles correlations between the learning and target concepts. Our recommendation policy is optimized by policy gradient. In addition, we also introduce an auxiliary module based on knowledge tracing to enhance the model’s stability by evaluating students’ learning effects on learning concepts. We conduct extensive experiments on two real-world public datasets and one industrial dataset, and the experimental results demonstrate the superiority and effectiveness of SRC. Code now is available at https://gitee.com/mindspore/models/tree/master/research/recommend/SRC.

Modeling multi-variate time-series (MVTS) data is a long-standing research subject and has found wide applications. Recently, there is a surge of interest in modeling spatial relations between variables as graphs, i.e., first learning one static graph for each dataset and then exploiting the graph structure via graph neural networks. However, as spatial relations may differ substantially across samples, building one static graph for all the samples inherently limits flexibility and severely degrades the performance in practice. To address this issue, we propose a framework for fine-grained modeling and utilization of spatial correlation between variables. By analyzing the statistical properties of real-world datasets, a universal decomposition of spatial correlation graphs is first identified. Specifically, the hidden spatial relations can be decomposed into a prior part, which applies across all the samples, and a dynamic part, which varies between samples, and building different graphs is necessary to model these relations. To better coordinate the learning of the two relational graphs, we propose a min-max learning paradigm that not only regulates the common part of different dynamic graphs but also guarantees spatial distinguishability among samples. The experimental results show that our proposed model outperforms the state-of-the-art baseline methods on both time-series forecasting and time-series point prediction tasks.

\AcIR is one of the most important fields to help users find relevant information. The interaction between IR systems and users can be naturally formulated as a decision-making problem. In the last decade, deep reinforcement learning (DRL) has become a promising direction to utilize the high model capacity of deep learning to improve long-term gains. On the one hand, there have been emerging research works focusing on leveraging DRL for IR tasks while the fundamental information theory under DRL settings, the principle of RL methods for IR tasks, or the experimental evaluation protocols of DRL-based IR systems, has not been deeply investigated. On the other hand, the emerging ChatGPT also provides new insights and challenges for DRL-based IR.

Learning-to-rank is a core technique in the top-N recommendation task, where an ideal ranker would be a mapping from an item set to an arrangement (a.k.a. permutation). Most existing solutions fall in the paradigm of probabilistic ranking principle (PRP), i.e., first score each item in the candidate set and then perform a sort operation to generate the top ranking list. However, these approaches neglect the contextual dependence among candidate items during individual scoring, and the sort operation is non-differentiable. To bypass the above issues, we propose Set-To-Arrangement Ranking (STARank), a new framework directly generates the permutations of the candidate items without the need for individually scoring and sort operations; and is end-to-end differentiable. As a result, STARank can operate when only the ground-truth permutations are accessible without requiring access to the ground-truth relevance scores for items. For this purpose, STARank first reads the candidate items in the context of the user browsing history, whose representations are fed into a Plackett-Luce module to arrange the given items into a list. To effectively utilize the given ground-truth permutations for supervising STARank, we leverage the internal consistency property of Plackett-Luce models to derive a computationally efficient list-wise loss. Experimental comparisons against 9 the state-of-the-art methods on 2 learning-to-rank benchmark datasets and 3 top-N real-world recommendation datasets demonstrate the superiority of STARank in terms of conventional ranking metrics. Notice that these ranking metrics do not consider the effects of the contextual dependence among the items in the list, we design a new family of simulation-based ranking metrics, where existing metrics can be regarded as special cases. STARank can consistently achieve better performance in terms of PBM and UBM simulation-based metrics.

Goal-oriented Learning path recommendation aims to recommend learning items (concepts or exercises) step-by-step to a learner to promote the mastery level of her specific learning goals. By formulating this task as a Markov decision process, reinforcement learning (RL) methods have demonstrated great power. Although extensive research efforts have been made, previous methods still fail to recommend effective goal-oriented paths due to the under-utilizing of goals. Specifically, it is mainly reflected in two aspects: (1)The lack of goal planning. When learners have multiple goals with different difficulties, the previous methods can't fully utilize the difficulties and dependencies between goal learning items to plan the sequence of achieving these goals, making the path chaotic and inefficient; (2)The lack of efficiency in goal achieving. When pursuing a single goal, the path may contain learning items unrelated to the goal, which makes realizing a certain goal inefficient. To address these challenges, we present a novel Graph Enhanced Hierarchical Reinforcement Learning (GEHRL) framework for goal-oriented learning path recommendation. The framework divides learning path recommendation into two parts: sub-goal selection(planning) and sub-goal achieving(learning item recommendation). Specifically, we employ a high-level agent as a sub-goal selector to select sub-goals for the low-level agent to achieve. The low-level agent in the framework is to recommend learning items to the learner. To make the path only contain goal-related learning items to improve the efficiency of achieving the goal, we develop a graph-based candidate selector to constrain the action space of the low-level agent based on the sub-goal and knowledge graph. We also develop test-based internal reward for low-level training so that the sparsity problem of external reward can be alleviated. Extensive experiments on three different simulators demonstrate our framework achieves state-of-the-art performance.

Position auctions are widely studied in the context of sponsored search advertising, where multiple ad slots are sold in a single auction. In traditional sponsored search, bids are submitted at the keyword level, while recent works have explored transitioning to impression-level bidding using Real-Time Bidding (RTB) techniques to achieve finer bidding. However, position auctions introduce varying user appeal across different positions and more dynamic auction landscape, which RTB, originally devised for single-slot display advertising, fails to address adequately.

Real-world data usually exhibits a long-tailed distribution, with a few frequent labels and a lot of few-shot labels. The study of institution name normalization is a perfect application case showing this phenomenon: there are many institutions worldwide, with enormous variations of their names in the publicly available literature. In this work, we first collect a large-scale institution name normalization dataset containing over 25k classes whose frequencies are naturally long-tail distributed. We construct our test set from four different subsets: many-, medium-, and few-shot sets, as well as a zero-shot open set, which are meant to isolate the few-shot and zero-shot learning scenarios from the massive many-shot classes. We also replicate several important benchmarks on our data, covering a wide range from search-based methods to neural network methods. Further, we propose our specially pretrained, BERT-based model that shows better out-of-distribution generalization on few-shot and zero-shot test sets. Compared to other datasets focusing on the long-tailed phenomenon, our dataset has one order of magnitude more training data than the largest existing long-tailed datasets and is naturally long-tailed rather than manually synthesized. We believe it provides an important and different scenario to study this problem. To our best knowledge, this is the first natural language dataset that focuses on this long-tailed and open-set classification problem.1

Re-ranking, as the final stage of the multi-stage recommender systems (MRS), aims at modeling the listwise context and the cross-item interactions between the candidate items. The objective is usually the overall utility (e.g., total clicks or revenue) of the re-ranked list, which is determined not only by the relevance, but also by the diversity of the list. However, existing methods equally promote diversity for all users and often compromise the relevance ranking. In reality, users have different diversity preferences and we should diversify the list tailored to individual users’ interests and needs. Users’ behavior history contains rich information which may be used for inferring their diversity preferences, but has rarely been explored in existing work. In this work, we propose a novel neural re-ranking with personalized diversification method (dubbed RAPID) to address the above challenge. RAPID explicitly models each user’s preference distribution over different topics by exploiting the intra- and inter-topic interactions from the user’s behavior history. The personalized diversity gain brought by each candidate item is then measured by the item’s marginal diversity and the learned personalized preference. The relevance and the personalized diversity are jointly optimized in an end-to-end manner to automatically manage the relevance-diversity tradeoff. Experimental results on two public datasets and a proprietary dataset show that RAPID outperforms the state-of-the-art with the highest utility and the best relevance-diversity tradeoff. We further prove that RAPID has a regret bound of $\tilde O(\sqrt n )$ on utility, which provides theoretical guarantee that its performance is near-optimal.

This paper introduces a distributed, GPU-centric experience replay system, GEAR, designed to perform scalable reinforcement learning (RL) with large sequence models (such as transformers). With such models, existing systems such as Reverb face considerable bottlenecks in memory, computation, and communication. GEAR, however, optimizes memory efficiency by enabling the memory resources on GPU servers (including host memory and device memory) to manage trajectory data. Furthermore, it facilitates decentralized GPU devices to expedite various trajectory selection strategies, circumventing computational bottlenecks. GEAR is equipped with GPU kernels capable of collecting trajectories using zero-copy access to host memory, along with remote-directed-memory access over InfiniBand, improving communication efficiency. Cluster experiments have shown that GEAR can achieve performance levels up to 6× greater than Reverb when training state-of-the-art large RL models. GEAR is open-sourced at https:// github.com/bigrl-team/gear.

Deep reinforcement learning has recently emerged as an appealing alternative for legged locomotion over multiple terrains by training a policy in physical simulation and then transferring it to the real world (i.e., sim-to-real transfer). Despite considerable progress, the capacity and scalability of traditional neural networks are still limited, which may hinder their applications in more complex environments. In contrast, the Transformer architecture has shown its superiority in a wide range of large-scale sequence modeling tasks, including natural language processing and decision-making problems. In this paper, we propose Terrain Transformer (TERT), a high-capacity Transformer model for quadrupedal locomotion control on various terrains. Furthermore, to better leverage Transformer in sim-to-real scenarios, we present a novel two-stage training framework consisting of an offline pretraining stage and an online correction stage, which can naturally integrate Transformer with privileged training. Extensive experiments in simulation demonstrate that TERT outperforms state-of-the-art baselines on different terrains in terms of return, energy consumption and control smoothness. In further real-world validation, TERT successfully traverses nine challenging terrains, including sand pit and stair down, which can not be accomplished by strong baselines.

Robots are traditionally bounded by a fixed embodiment during their operational lifetime, which limits their ability to adapt to their surroundings. Co-optimizing control and morphology of a robot, however, is often inefficient due to the complex interplay between the controller and morphology. In this paper, we propose a learning-based control method that can inherently take morphology into consideration such that once the control policy is trained in the simulator, it can be easily deployed to real robots with different embodiments. In particular, we present the Embodiment-aware Transformer (EAT), an architecture that casts this control problem as conditional sequence modeling. EAT outputs the optimal actions by leveraging a causally masked Transformer. By conditioning an autoregressive model on the desired robot embodiment, past states, and actions, our EAT model can generate future actions that best fit the current robot embodiment. Experimental results show that EAT can outperform all other alternatives in embodiment-varying tasks, and succeed in an example of real-world evolution tasks: stepping down a stair through updating the morphology alone. We hope that EAT will inspire a new push toward real-world evolution across many domains, where algorithms like EAT can blaze a trail by bridging the field of evolutionary robotics and big data sequence modeling.

Over recent decades, sequential decision-making tasks are mostly tackled with expert systems and reinforcement learning. However, these methods are still incapable of being generalizable enough to solve new tasks at a low cost. In this article, we discuss a novel paradigm that leverages Transformer-based sequence models to tackle decision-making tasks, named large decision models. Starting from offline reinforcement learning scenarios, early attempts demonstrate that sequential modeling methods can be applied to train an effective policy given sufficient expert trajectories. When the sequence model goes large, its generalization ability over a variety of tasks and fast adaptation to new tasks has been observed, which is highly potential to enable the agent to achieve artificial general intelligence for sequential decision-making in the near future.

Order execution is a fundamental task in quantitative finance, aiming at finishing acquisition or liquidation for a number of trading orders of the specific assets. Recent advance in model-free reinforcement learning (RL) provides a data-driven solution to the order execution problem. However, the existing works always optimize execution for an individual order, overlooking the practice that multiple orders are specified to execute simultaneously, resulting in suboptimality and bias. In this paper, we first present a multi-agent RL (MARL) method for multi-order execution considering practical constraints. Specifically, we treat every agent as an individual operator to trade one specific order, while keeping communicating with each other and collaborating for maximizing the overall profits. Nevertheless, the existing MARL algorithms often incorporate communication among agents by exchanging only the information of their partial observations, which is inefficient in complicated financial market. To improve collaboration, we then propose a learnable multi-round communication protocol, for the agents communicating the intended actions with each other and refining accordingly. It is optimized through a novel action value attribution method which is provably consistent with the original learning objective yet more efficient. The experiments on the data from two real-world markets have illustrated superior performance with significantly better collaboration effectiveness achieved by our method.

With the widespread application of online advertising systems, click-through rate (CTR) prediction has received more and more attention and research. The most prominent features of CTR prediction are its multi-field categorical data format, and vast and daily-growing data volume (e.g., billions of user click logs). The large capacity of neural models helps digest such massive amounts of data under the supervised learning paradigm, yet they fail to utilize the substantial data to its full potential, since click signals are not sufficient enough for the model to learn capable representations of features and instances. The self-supervised learning paradigm provides a more promising pretrain-finetune solution to better exploit the large amount of user click logs and learn more robust and effective representations. However, current works on this line are still preliminary and rudimentary, leaving self-supervised learning for CTR prediction still an open question. To this end, we propose a Model-agnostic Pretraining (MAP) framework that applies feature corruption and recovery on multi-field categorical data, and more specifically, we derive two practical algorithms: masked feature prediction (MFP) and replaced feature detection (RFD). MFP digs into feature interactions within each instance through masking and predicting a small portion of input features, and we also introduce Noise Contrastive Estimation (NCE) to handle large feature spaces. RFD further turns MFP into a binary classification mode through replacing and detecting changes in input features, making it even simpler and more effective for CTR pretraining. Our extensive experiments on two real-world million-level datasets (i.e., Avazu, Criteo) demonstrate the advantages of these two methods over several strong baselines, and achieve new state-of-the-art in terms of both performance and efficiency for CTR prediction.

Real-Time Bidding (RTB) has shown remarkable success in display advertising and has been employed in other advertising scenarios, e.g., sponsored search advertising with multiple ad slots. Many current RTB techniques built for single-slot display advertising are thus no longer applicable, especially in the bid landscape forecasting. Landscape forecasting predicts market competition, including the highest bid price and winning probability, which is preliminary and crucial for the subsequent bidding strategy design. In the multi-slot advertising, predicting the winning prices for each position requires a more precise differentiation of bids among top advertisers. Furthermore, defining the winning probability and addressing censorship issues are not as straightforward as in the case of a single slot. In view of these challenges, how to forecast the bidding landscape in the multi-slot environment remains open.

Computerized Adaptive Testing (CAT) refers to an online system that adaptively selects the best-suited question for students with various abilities based on their historical response records. Compared with traditional CAT methods based on heuristic rules, recent data-driven CAT methods obtain higher performance by learning from large-scale datasets. However, most CAT methods only focus on the quality objective of predicting the student ability accurately, but neglect concept diversity or question exposure control, which are important considerations in ensuring the performance and validity of CAT. Besides, the students' response records contain valuable relational information between questions and knowledge concepts. The previous methods ignore this relational information, resulting in the selection of sub-optimal test questions. To address these challenges, we propose a Graph-Enhanced Multi-Objective method for CAT (GMOCAT). Firstly, three objectives, namely quality, diversity and novelty, are introduced into the Scalarized Multi-Objective Reinforcement Learning framework of CAT, which respectively correspond to improving the prediction accuracy, increasing the concept diversity and reducing the question exposure. We use an Actor-Critic Recommender to select questions and optimize three objectives simultaneously by the scalarization function. Secondly, we utilize the graph neural network to learn relation-aware embeddings of questions and concepts. These embeddings are able to aggregate neighborhood information in the relation graphs between questions and concepts. We conduct experiments on three real-world educational datasets. The experimental results show that GMOCAT not only outperforms the state-of-the-art methods in the ability prediction, but also achieve superior performance in improving the concept diversity and alleviating the question exposure.

As an emerging field driven by industrial applications, integrated re-ranking combines lists from upstream sources into a single list, and presents it to the user. The quality of integrated re-ranking is especially sensitive to real-time user behaviors and preferences. However, existing methods are all built on the cloud-to-edge framework, where mixed lists are generated by the cloud model and then sent to the devices. Despite its effectiveness, such a framework fails to capture users' real-time preferences due to the network bandwidth and latency. Hence, we propose to place the integrated re-ranking model on devices, allowing for the full exploitation of real-time behaviors. To achieve this, we need to address two key issues: first, how to extract users' preferences for different sources from heterogeneous and imbalanced user behaviors; second, how to explore the correlation between the extracted personalized preferences and the candidate items. In this work, we present the first on-Device Integrated Re-ranking framework, DIR, to avoid delays in processing real-time user behaviors. DIR includes a multi-sequence behavior modeling module to extract the user's source-level preferences, and a preference-adaptive re-ranking module to incorporate personalized source-level preferences into the re-ranking of candidate items. Besides, we design exposure loss and utility loss to jointly optimize exposure fairness and overall utility. Extensive experiments on three datasets show that DIR significantly outperforms the state-of-the-art baselines in utility-based and fairness-based metrics.

Learning representations purely from observations concerns the problem of learning a low-dimensional, compact representation which is beneficial to prediction models. Under the hypothesis that the intrinsic latent factors follow some casual generative models, we argue that by learning a causal representation, which is the minimal sufficient causes of the whole system, we can improve the robustness and generalization performance of machine learning models. In this paper, we develop a learning method to learn such representation from observational data by regularizing the learning procedure with mutual information measures, according to the hypothetical factored causal graph. We theoretically and empirically show that the models trained with the learned causal representations are more robust under adversarial attacks and distribution shifts compared with baselines.

Data science is concerned with mining data patterns from a database, which is assembled by tabular data. As the routine of machine learning, most of the previous work mining the tabular data's pattern based on a single instance. However, they neglect the similar tabular data instances that could help make the label prediction of the target data instance. Recently, some retrieval-based methods for tabular data label prediction have been proposed, which, however, treat the data as sparse vectors to perform the retrieval, which fails to make use of the semantic information of the tabular data. To address such a problem, in this paper, we propose a novel framework of dense retrieval on tabular data (DERT) to support flexible data representation learning and effective label prediction on tabular data. DERT consists of two major components: (i) the encoder that makes the tabular data as embeddings, which could be trained by flexible neural networks and auxiliary loss functions; (ii) the retrieval and prediction component, which makes use of similar rows in the table to make label prediction of the target row. We test DERT on two tasks based on five real-world datasets and experimental results show that DERT achieves consistent improvements over the state-of-the-art and various baselines.

Industrial recommender systems face the challenge of operating in non-stationary environments, where data distribution shifts arise from evolving user behaviors over time. To tackle this challenge, a common approach is to periodically re-train or incrementally update deployed deep models with newly observed data, resulting in a continual learning process. However, the conventional learning paradigm of neural networks relies on iterative gradient-based updates with a small learning rate, making it slow for large recommendation models to adapt. In this paper, we introduce ReLoop2, a self-correcting learning loop that facilitates fast model adaptation in online recommender systems through responsive error compensation. Inspired by the slow-fast complementary learning system observed in human brains, we propose an error memory module that directly stores error samples from incoming data streams. These stored samples are subsequently leveraged to compensate for model prediction errors during testing, particularly under distribution shifts. The error memory module is designed with fast access capabilities and undergoes continual refreshing with newly observed data samples during the model serving phase to support fast model adaptation. We evaluate the effectiveness of ReLoop2 on three open benchmark datasets as well as a real-world production dataset. The results demonstrate the potential of ReLoop2 in enhancing the responsiveness and adaptiveness of recommender systems operating in non-stationary environments.

User-tag profiling is an effective way of mining user attributes in modern recommender systems. However, prior researches fail to extract users' precise preferences for tags in the items due to their incomplete feature-input patterns. To convert user-item interactions to user-tag preferences, we propose a novel feature-based framework named Coalition Tag Multi-View Mapping (CTMVM), which identifies and investigates two special features, Coalition Feature and Privileged Feature. The former indicates decisive tags in each click where relationships between tags in one item are treated as a coalition game. The latter represents highly informative features that only occur during training. For the coalition feature, we adopt Shapley Value based Empowerment (SVE) to model the tags in items with a game-theoretic paradigm and charge the network to straight master user preferences for essential tags. For the privileged feature, we present Privileged Knowledge Mapping (PKM) to explicitly distill privileged feature knowledge for each tag into one single embedding, which assists the model in predicting user-tag preferences at a more fine-grained level. However, the barren capacity of single embeddings limits the diverse relations between each tag and different privileged features. Therefore, we further propose Adaptive Multi-View Mapping (AMVM) model to enhance effect by handling multiple mapping networks. Excellent offline experiment results on two public and one private datasets show the out-standing performance of CTMVM. After the deployment on Alibaba large-scale recommendation systems, CTMVM achieved improvement by 10.81% and 6.74% in terms of Theme-CTR and Item-CTR respectively, which validates the effectiveness of taking in the two particular features for training.

Diffusion models have demonstrated highly-expressive generative capabilities in vision and NLP. Recent studies in reinforcement learning (RL) have shown that diffusion models are also powerful in modeling complex policies or trajectories in offline datasets. However, these works have been limited to single-task settings where a generalist agent capable of addressing multi-task predicaments is absent. In this paper, we aim to investigate the effectiveness of a single diffusion model in modeling large-scale multi-task offline data, which can be challenging due to diverse and multimodal data distribution. Specifically, we propose Multi-Task Diffusion Model (\textsc{MTDiff}), a diffusion-based method that incorporates Transformer backbones and prompt learning for generative planning and data synthesis in multi-task offline settings. \textsc{MTDiff} leverages vast amounts of knowledge available in multi-task data and performs implicit knowledge sharing among tasks. For generative planning, we find \textsc{MTDiff} outperforms state-of-the-art algorithms across 50 tasks on Meta-World and 8 maps on Maze2D. For data synthesis, \textsc{MTDiff} generates high-quality data for testing tasks given a single demonstration as a prompt, which enhances the low-quality datasets for even unseen tasks.

An intelligent conversational agent (a.k.a., chat-bot) could embrace conversational technologies to obtain user preferences online, to overcome inherent limitations of recommender systems trained over the offline historical user behaviors. In this paper, we propose CORE, a new offline-training and online-checking framework to plug a COnversational agent into REcommender systems. Unlike most prior conversational recommendation approaches that systemically combine conversational and recommender parts through a reinforcement learning framework, CORE bridges the conversational agent and recommender system through a unified uncertainty minimization framework, which can be easily applied to any existing recommendation approach. Concretely, CORE treats a recommender system as an offline estimator to produce an estimated relevance score for each item, while CORE regards a conversational agent as an online checker that checks these estimated scores in each online session. We define uncertainty as the sum of unchecked relevance scores. In this regard, the conversational agent acts to minimize uncertainty via querying either attributes or items. Towards uncertainty minimization, we derive the certainty gain of querying each attribute and item, and develop a novel online decision tree algorithm to decide what to query at each turn. Our theoretical analysis reveals the bound of the expected number of turns of CORE in a cold-start setting. Experimental results demonstrate that CORE can be seamlessly employed on a variety of recommendation approaches, and can consistently bring significant improvements in both hot-start and cold-start settings.

extended-abstract Share on International Workshop on Deep Learning Practice for High-Dimensional Sparse Data with RecSys 2023 Authors: Ruiming Tang Huawei Noah's Ark Lab, China Huawei Noah's Ark Lab, China 0000-0002-9224-2431View Profile , Xiaoqiang Zhu Mobvista Group, China Mobvista Group, China 0000-0001-7486-0853View Profile , Junfeng Ge Alibaba Group, China Alibaba Group, China 0000-0001-8435-0443View Profile , Kuang-chih Lee Alibaba Group, USA Alibaba Group, USA 0009-0007-5198-9866View Profile , Biye Jiang Alibaba Group, China Alibaba Group, China 0009-0001-5814-1581View Profile , Xingxing Wang Meituan, China Meituan, China 0000-0002-2655-3928View Profile , Han Zhu Alibaba Group, China Alibaba Group, China 0000-0002-9522-5637View Profile , Tao Zhuang Alibaba Group, China Alibaba Group, China 0000-0002-7408-8514View Profile , Weiwen Liu Huawei Noah's Ark Lab, China Huawei Noah's Ark Lab, China 0000-0002-9148-3997View Profile , Kan Ren Microsoft Research, China Microsoft Research, China 0000-0002-4032-9615View Profile , Weinan Zhang Shanghai Jiao Tong University, China Shanghai Jiao Tong University, China 0000-0002-0127-2425View Profile , Xiangyu Zhao City University of Hong Kong, China City University of Hong Kong, China 0000-0003-2926-4416View Profile Authors Info & Claims RecSys '23: Proceedings of the 17th ACM Conference on Recommender SystemsSeptember 2023Pages 1276–1280https://doi.org/10.1145/3604915.3608765Published:14 September 2023Publication History 0citation67DownloadsMetricsTotal Citations0Total Downloads67Last 12 Months67Last 6 weeks67 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

Most click models focus on user behaviors towards a single list. However, with the development of user interface (UI) design, the layout of displayed items on a result page tends to be multi-block style instead of a single list, which requires different assumptions to model user behaviors more accurately. There exist click models for multi-block pages in desktop contexts, but they cannot be directly applied to mobile scenarios due to different interaction manners, result types and especially multi-block presentation styles. In particular, multi-block mobile pages can normally be decomposed into interleavings of basic vertical blocks and horizontal blocks, thus resulting in typically F-shape forms. To mitigate gaps between desktop and mobile contexts for multi-block pages, we conduct a user eye-tracking study, and identify users' sequential browsing, block skip and comparison patterns on F-shape pages. These findings lead to the design of a novel F-shape Click Model (FSCM), which serves as a general solution to multi-block mobile pages. Firstly, we construct a Directed Acyclic Graph (DAG) for each page, where each item is regarded as a vertex and each edge indicates the user's possible examination flow. Secondly, we propose DAG-structured GRUs and a comparison module to model users' sequential (sequential browsing, block skip) and non-sequential (comparison) behaviors respectively. Finally, we combine GRU states and comparison patterns to perform user click predictions. Experiments show that FSCM outperforms baseline models.

This paper is concerned with TV recommendation, where one major challenge is the coupling behavior issue that the behaviors of multiple users are coupled together and not directly distinguishable because the users share the same account. Unable to identify the current watching user and use the coupling behaviors directly could lead to sub-optimal recommendation results due to the noise introduced by the behaviors of other users. Most existing methods deal with this issue either by unsupervised clustering algorithms or depending on latent user representation learning with strong assumptions. However, they neglect to sophisticatedly model the current session behaviors, which carry the information of user identification. Another critical limitation of the existing models is the lack of supervision signal on distinguishing behaviors because they solely depend on the final click label, which is insufficient to provide effective supervision. To address the above problems, we propose the Coupling Sequence Model (COSMO) for TV recommendation. In COSMO, we design a session-aware co-attention mechanism that uses both the candidate item and session behaviors as the query to attend to the historical behaviors in a fine-grained manner. Furthermore, we propose to use the data of accounts with multiple devices (e.g., families with various TV sets), which means the behaviors of one account are generated on different devices. We regard the device information as weak supervision and propose a novel pair-wise attention loss for learning to distinguish the coupling behaviors. Extensive offline experiments and online A/B tests over a commercial TV service provider demonstrate the efficacy of COSMO compared to the existing models.

Reranking, as the final stage of multi-stage recommender systems, refines the initial lists to maximize the total utility. With the development of multimedia and user interface design, the recommendation page has evolved to a multi-list style. Separately employing traditional list-level reranking methods for different lists overlooks the inter-list interactions and the effect of different page formats, thus yielding suboptimal reranking performance. Moreover, simply applying a shared network for all the lists fails to capture the commonalities and distinctions in user behaviors on different lists. To this end, we propose to draw a bird's-eye view of page-level reranking and design a novel Page-level Attentional Reranking (PAR) model. We introduce a hierarchical dual-side attention module to extract personalized intra- and inter-list interactions. A spatial-scaled attention network is devised to integrate the spatial relationship into pairwise item influences, which explicitly models the page format. The multi-gated mixture-of-experts module is further applied to capture the commonalities and differences of user behaviors between different lists. Extensive experiments on a public dataset and a proprietary dataset show that PAR significantly outperforms existing baseline models.

Considering the balance between revenue and resource consumption for industrial recommender systems, intelligent recommendation computing has been emerging recently. Existing solutions deploy the same recommendation model to serve users indiscriminately, which is sub-optimal for total revenue maximization. We propose a multi-model service solution by deploying different-complexity models to serve different-valued users. An automated dynamic model generation framework AutoGen is elaborated to efficiently derive multiple parameter-sharing models with diverse complexities and adequate predictive capabilities. A mixed search space is designed and an importance-aware progressive training scheme is proposed to prevent interference between different architectures, which avoids the model retraining and improves the search efficiency, thereby efficiently deriving multiple models. Extensive experiments are conducted on two public datasets to demonstrate the effectiveness and efficiency of AutoGen.

Compared to model-free reinforcement learning (RL), model-based RL is often more sample efficient by leveraging a learned dynamics model to help decision making. However, the learned model is usually not perfectly accurate and the error will compound in multi-step predictions, which can lead to poor asymptotic performance. In this paper, we first derive an upper bound of the return discrepancy between the real dynamics and the learned model, which reveals the fundamental problem of distribution shift between simulated data and real data. Inspired by the theoretical analysis, we propose an adaptation augmented model-based policy optimization (AMPO) framework to address the distribution shift problem from the perspectives of feature learning and instance re-weighting, respectively. Specifically, the feature-based variant, namely FAMPO, introduces unsupervised model adaptation to minimize the integral probability metric (IPM) between feature distributions from real and simulated data, while the instance-based variant, termed as IAMPO, utilizes importance sampling to re-weight the real samples used to train the model. Besides model learning, we also investigate how to improve policy optimization in the model usage phase by selecting simulated samples with different probability according to their uncertainty. Extensive experiments on challenging continuous control tasks show that FAMPO and IAMPO, coupled with our model usage technique, achieves superior performance against baselines, which demonstrates the effectiveness of the proposed methods.

Population-based multi-agent reinforcement learning (PB-MARL) encompasses a range of methods that merge dynamic population selection with multi-agent reinforcement learning algorithms (MARL). While PB-MARL has demonstrated notable achievements in complex multi-agent tasks, its sequential execution is plagued by low computational efficiency due to the diversity in computing patterns and policy combinations. We propose a solution involving a stateless central task dispatcher and stateful workers to handle PB-MARL's subroutines, thereby capitalizing on parallelism across various components for efficient problem-solving. In line with this approach, we introduce MALib, a parallel framework that incorporates a task control model, independent data servers, and an abstraction of MARL training paradigms. The framework has undergone extensive testing and is available under the MIT license (https://github.com/sjtu-marl/malib)

Although reinforcement learning (RL) techniques are regarded as promising solutions for interactive recommender systems (IRS), such solutions still face three main challenges, namely, i) time inefficiency when handling large discrete action space in IRS, ii) inability to deal with the cold-start scenarios in IRS, iii) data inefficiency during training the RL-based methods. To tackle these challenges, we propose a generic tree-structured RL framework taking both policy-based and value-based approaches into consideration. We propose to construct a balanced tree over representations of the items, such that picking an item is formulated as seeking a suitable path from the root to a leaf node in the balanced tree, which dramatically reduces the time complexity of item recommendation. Further, for cold-start scenarios where prior information of the items is unavailable, we initialize a random balanced tree as the starting point and then refine the tree structure based on the learned item representations. Besides, we also incorporate a user modeling component to explicitly model the environment, which can be utilized in the training phase to improve data efficiency. Extensive experiments on two real-world datasets are conducted and demonstrate that our framework can achieve superior recommendation performance and provide time and data efficiency improvement over state-of-the-art methods in both warm-start and cold-start IRS scenarios.

Feature embedding learning and feature interaction modeling are two crucial components of deep models for Click-Through Rate (CTR) prediction in recommender systems. Most existing deep CTR models suffer from the following three problems. First, feature interactions are either manually designed or simply enumerated. However, not all the feature interactions are useful for the prediction task and useless feature interactions may introduce noisy signals thus causing overfitting. Second, all the feature interactions are modeled with an identical interaction function, whereas different interaction functions introduce different inductive biases to better capture various feature interaction patterns. Third, in most existing models, different features share the same embedding size. However, model size can be further optimized without sacrificing performance by differentiating embedding sizes for individual features, as the amount of information contained in each feature varies much. To address the three issues mentioned above, we proposeAutomatic Interaction Machine (AIM)with three core components, namely, Feature Interaction Search (FIS), Interaction Function Search (IFS) and Embedding Dimension Search (EDS), respectively. To tackle the first problem, FIS component automatically identifies different orders of essential feature interactions with useless ones pruned. Taking care of the second problem, IFS component selects appropriate interaction functions for each individual feature interaction in a learnable way. Moreover, to avoid learning conflict among different interaction functions, IFS proposes function-wise embeddings via performing multiple embeddings for each feature, where each feature embedding corresponds to one possible interaction function. However, utilizing multiple embeddings for each feature may make the model size affordably large if we keep the same embedding size as utilizing shared embedding (i.e., each feature shares the same embedding for different interaction functions). To solve this third problem, EDS automatically selects proper embedding size for each feature. Such a flexible embedding size adaptation is able to reduce the large amount of embedding parameters introduced by function-wise embeddings. Offline experiments on three large-scale datasets (two public benchmarks, one private dataset) validate that AIM can significantly improve various FM-based models. AIM has been deployed in the recommendation service of a mainstream app market, where a three-week online A/B test demonstrated the superiority of AIM, improving DeepFM model by 4.4% in terms of CTR.

Click-through rate (CTR) estimation plays a crucial role in modern online personalization services. It is essential to capture users’ drifting interests by modeling sequential user behaviors to build an accurate CTR estimation model. However, as the users accumulate a large amount of behavioral data on the online platforms, the current CTR models have to truncate user behavior sequences and utilize the most recent behaviors, which leads to a problem that sequential patterns such as periodicity or long-term dependency are not contained in the recent behaviors but in far back history. However, it is non-trivial to model the entire user sequence by directly using it for two reasons. Firstly, the very long input sequences will make online inference time and system load infeasible. Secondly, the very long sequences contain much noise, thus making it difficult for CTR models to capture useful patterns effectively. To tackle this issue, we consider it from the input data perspective instead of designing more sophisticated yet complex models. As the entire user behavior sequence contains much noise, it is unnecessary to input the entire sequence. Instead, we could just retrieve only a small part of it as the input to the CTR model. In this article, we propose the U ser B ehavior R etrieval (UBR) framework which aims at learning to retrieve the most informative user behaviors according to each CTR estimation request. Retrieving only a small set of behaviors could alleviate the two problems of utilizing very long sequences (i.e., inference efficiency and noisy input). The distinguishing property of UBR is that it supports arbitrary and learnable retrieval functions instead of utilizing a fixed pre-defined function, which is different from the current retrieval-based methods. Offline evaluations on three large-scale real-world datasets demonstrate the superiority and efficacy of the UBR framework. We further deploy UBR at the Huawei App Store, where it achieves 6.6% of eCPM gain in the online A/B test and now serves the main traffic in the Huawei App Store advertising scenario.

Program representation, which aims at converting program source code into vectors with automatically extracted features, is a fundamental problem in programming language processing (PLP). Recent work tries to represent programs with neural networks based on source code structures. However, such methods often focus on the syntax and consider only one single perspective of programs, limiting the representation power of models. This paper proposes a multi-view graph (MVG) program representation method. MVG pays more attention to code semantics and simultaneously includes both data flow and control flow as multiple views. These views are then combined and processed by a graph neural network (GNN) to obtain a comprehensive program representation that covers various aspects. We thoroughly evaluate our proposed MVG approach in the context of algorithm detection, an important and challenging subfield of PLP. Specifically, we use a public dataset POJ-104 and also construct a new challenging dataset ALG-109 to test our method. In experiments, MVG outperforms previous methods significantly, demonstrating our model's strong capability of representing source code.

Span-based methods with the neural networks backbone have great potential for the nested named entity recognition (NER) problem.However, they face problems such as degenerating when positive instances and negative instances largely overlap.Besides, the generalization ability matters a lot in nested NER, as a large proportion of entities in the test set hardly appear in the training set.In this work, we try to improve the span representation by utilizing retrieval-based span-level graphs, connecting spans and entities in the training data based on n-gram features.Specifically, we build the entity-entity graph and span-entity graph globally based on n-gram similarity to integrate the information of similar neighbor entities into the span representation.To evaluate our method, we conduct experiments on three common nested NER datasets, ACE2004, ACE2005, and GENIA datasets.Experimental results show that our method achieves general improvements on all three benchmarks (+0.30∼ 0.85 micro-F1), and obtains special superiority on low frequency entities (+0.56 ∼ 2.08 recall).

Algorithmic decisions made by machine learning models in high-stakes domains may have lasting impacts over time. However, naive applications of standard fairness criterion in static settings over temporal domains may lead to delayed and adverse effects. To understand the dynamics of performance disparity, we study a fairness problem in Markov decision processes (MDPs). Specifically, we propose return parity, a fairness notion that requires MDPs from different demographic groups that share the same state and action spaces to achieve approximately the same expected time-discounted rewards. We first provide a decomposition theorem for return disparity, which decomposes the return disparity of any two MDPs sharing the same state and action spaces into the distance between group-wise reward functions, the discrepancy of group policies, and the discrepancy between state visitation distributions induced by the group policies. Motivated by our decomposition theorem, we propose algorithms to mitigate return disparity via learning a shared group policy with state visitation distributional alignment using integral probability metrics. We conduct experiments to corroborate our results, showing that the proposed algorithm can successfully close the disparity gap while maintaining the performance of policies on two real-world recommender system benchmark datasets.

Modelling the user's multiple behaviors is an essential part of modern e-commerce, whose widely adopted application is to jointly optimize click-through rate (CTR) and conversion rate (CVR) predictions. Most of existing methods overlook the effect of two key characteristics of the user's behaviors: for each item list, (i) contextual dependence refers to that the user's behaviors on any item are not purely determinated by the item itself but also are influenced by the user's previous behaviors (e.g., clicks, purchases) on other items in the same sequence; (ii) multiple time scales means that users are likely to click frequently but purchase periodically. To this end, we develop a new multi-scale user behavior network named H ierarchical r E current R anking O n the E ntire S pace (HEROES) which incorporates the contextual information to estimate the user multiple behaviors in a multi-scale fashion. Concretely, we introduce a hierarchical framework, where the lower layer models the user's engagement behaviors while the upper layer estimates the user's satisfaction behaviors. The proposed architecture can automatically learn a suitable time scale for each layer to capture the dynamic user's behavioral patterns. Besides the architecture, we also introduce the Hawkes process to form a novel recurrent unit which can not only encode the items' features in the context but also formulate the excitation or discouragement from the user's previous behaviors. We further show that HEROES can be extended to build unbiased ranking systems through combinations with the survival analysis technique. Extensive experiments over three large-scale industrial datasets demonstrate the superiority of our model compared with the state-of-the-art methods. characteristics of the user's behaviors: for each item list, (i) contex- tual dependence refers to that the user's behaviors on any item are not purely determinated by the item itself but also are influenced by the user's previous behaviors (e.g., clicks, purchases) on other items in the same sequence; (ii) multiple time scales means that users are likely to click frequently but purchase periodically. To this end, we develop a new multi-scale user behavior network named Hierarchical rEcurrent Ranking On the Entire Space (HEROES) which incorporates the contextual information to estimate the user multiple behaviors in a multi-scale fashion. Concretely, we intro- duce a hierarchical framework, where the lower layer models the user's engagement behaviors while the upper layer estimates the user's satisfaction behaviors. The proposed architecture can auto- matically learn a suitable time scale for each layer to capture the dynamic user's behavioral patterns. Besides the architecture, we also introduce the Hawkes process to form a novel recurrent unit which can not only encode the items' features in the context but also formulate the excitation or discouragement from the user's previous behaviors. We further show that HEROES can be extended to build unbiased ranking systems through combinations with the survival analysis technique. Extensive experiments over three large- scale industrial datasets demonstrate the superiority of our model compared with the state-of-the-art methods.

While end-to-end neural machine translation (NMT) has achieved impressive progress, noisy input usually leads models to become fragile and unstable. Generating adversarial examples as the augmented data has been proved to be useful to alleviate this problem. Existing methods for adversarial example generation (AEG) are word-level or character-level, which ignore the ubiquitous phrase structure. In this paper, we propose a Phrase-level Adversarial Example Generation (PAEG) framework to enhance the robustness of the translation model. Our method further improves the gradient-based word-level AEG method by adopting a phrase-level substitution strategy. We verify our method on three benchmarks, including LDC Chinese-English, IWSLT14 German-English, and WMT14 English-German tasks. Experimental results demonstrate that our approach significantly improves translation performance and robustness to noise compared to previous strong baselines.

Recent progress in state-only imitation learning extends the scope of applicability of imitation learning to real-world settings by relieving the need for observing expert actions. However, existing solutions only learn to extract a state-to-action mapping policy from the data, without considering how the expert plans to the target. This hinders the ability to leverage demonstrations and limits the flexibility of the policy. In this paper, we introduce Decoupled Policy Optimization (DePO), which explicitly decouples the policy as a high-level state planner and an inverse dynamics model. With embedded decoupled policy gradient and generative adversarial training, DePO enables knowledge transfer to different action spaces or state transition dynamics, and can generalize the planner to out-of-demonstration state regions. Our in-depth experimental analysis shows the effectiveness of DePO on learning a generalized target state planner while achieving the best imitation performance. We demonstrate the appealing usage of DePO for transferring across different tasks by pre-training, and the potential for co-training agents with various skills.

Quality Diversity (QD) algorithms in population-based reinforcement learning aim to optimize agents' returns and diversity among the population simultaneously. It is conducive to solving exploration problems in reinforcement learning and potentially getting multiple good and diverse strategies. However, previous methods typically define behavioral embedding in action space or outcome space, which neglect trajectory characteristics during the execution process. In this paper, we introduce a trajectory embedding model trained by Variational Autoencoder with similarity constraint to characterize trajectory features. Based on that, we propose a hierarchical trajectory-space exploration (HTSE) framework using Determinantal Point Processes (DPP) to generate high-quality and diverse solutions in the selection and mutation process. The experimental results show that our HTSE method effectively completes several simulated tasks, outperforming other Quality-Diversity Reinforcement Learning algorithms.

As the final stage of the multi-stage recommender system (MRS), re-ranking directly affects users’ experience and satisfaction by rearranging the input ranking lists, and thereby plays a critical role in MRS. With the advances in deep learning, neural re-ranking has become a trending topic and been widely adopted in industrial applications. This review aims at integrating re-ranking algorithms into a broader picture, and paving ways for more comprehensive solutions for future research. For this purpose, we first present a taxonomy of current methods on neural re-ranking. Then we give a description of these methods along with the historic development according to their objectives. The network structure, personalization, and complexity are also discussed and compared. Next, we provide a benchmark for the major neural re-ranking models and quantitatively analyze their re-ranking performance. Finally, the review concludes with a discussion on future prospects of this field. A list of papers discussed in this review, the benchmark datasets, our re-ranking library LibRerank, and detailed parameter settings are publicly available at https://github.com/LibRerank-Community/LibRerank.

Goal-conditioned reinforcement learning (GCRL), related to a set of complex RL problems, trains an agent to achieve different goals under particular scenarios. Compared to the standard RL solutions that learn a policy solely depending on the states or observations, GCRL additionally requires the agent to make decisions according to different goals. In this survey, we provide a comprehensive overview of the challenges and algorithms for GCRL. Firstly, we answer what the basic problems are studied in this field. Then, we explain how goals are represented and present how existing solutions are designed from different points of view. Finally, we make the conclusion and discuss potential future prospects that recent researches focus on.

It is challenging for reinforcement learning (RL) algorithms to succeed in real-world applications. Take financial trading as an example, the market information is noisy yet imperfect and the macroeconomic regulation or other factors may shift between training and evaluation, thus it requires both generalization and high sample efficiency for resolving the task. However, directly applying typical RL algorithms can lead to poor performance in such scenarios. To derive a robust and applicable RL algorithm, in this work, we design a simple but effective method named Ensemble Proximal Policy Optimization (EPPO), which learns ensemble policies in an end-to-end manner. Notably, EPPO combines each policy and the policy ensemble organically and optimizes both simultaneously. In addition, EPPO adopts a diversity enhancement regularization over the policy space which helps to generalize to unseen states and promotes exploration. We theoretically prove that EPPO can increase exploration efficacy, and through comprehensive experimental evaluations on various tasks, we demonstrate that EPPO achieves higher efficiency and is robust for real-world applications compared with vanilla policy optimization algorithms and other ensemble methods. Code and supplemental materials are available at https://seqml.github.io/eppo.

Spatio-temporal graph learning is a key method for urban computing tasks, such as traffic flow, taxi demand and air quality forecasting. Due to the high cost of data collection, some developing cities have few available data, which makes it infeasible to train a well-performed model. To address this challenge, cross-city knowledge transfer has shown its promise, where the model learned from data-sufficient cities is leveraged to benefit the learning process of data-scarce cities. However, the spatio-temporal graphs among different cities show irregular structures and varied features, which limits the feasibility of existing Few-Shot Learning (FSL) methods. Therefore, we propose a model-agnostic few-shot learning framework for spatio-temporal graph called ST-GFSL. Specifically, to enhance feature extraction by transferring cross-city knowledge, ST-GFSL proposes to generate non-shared parameters based on node-level meta knowledge. The nodes in target city transfer the knowledge via parameter matching, retrieving from similar spatio-temporal characteristics. Furthermore, we propose to reconstruct the graph structure during meta-learning. The graph reconstruction loss is defined to guide structure-aware learning, avoiding structure deviation among different datasets. We conduct comprehensive experiments on four traffic speed prediction benchmarks and the results demonstrate the effectiveness of ST-GFSL compared with state-of-the-art methods.

With the tremendous expansion of graphs data, node classification shows its great importance in many real-world applications. Existing graph neural network based methods mainly focus on classifying unlabeled nodes within fixed classes with abundant labeling. However, in many practical scenarios, graph evolves with emergence of new nodes and edges. Novel classes appear incrementally along with few labeling due to its newly emergence or lack of exploration. In this paper, we focus on this challenging but practical graph few-shot class-incremental learning (GFSCIL) problem and propose a novel method called Geometer. Instead of replacing and retraining the fully connected neural network classifier, Geometer predicts the label of a node by finding the nearest class prototype. Prototype is a vector representing a class in the metric space. With the pop-up of novel classes, Geometer learns and adjusts the attention-based prototypes by observing the geometric proximity, uniformity and separability. Teacher-student knowledge distillation and biased sampling are further introduced to mitigate catastrophic forgetting and unbalanced labeling problem respectively. Experimental results on four public datasets demonstrate that Geometer achieves a substantial improvement of 9.46% to 27.60% over state-of-the-art methods.

Recently, we have witnessed that deep learning-based approaches have been widely applied. Particularly, some applications involve data that are high dimensional, sparse or imbalanced, which are different from those applications with dense data processing, such as image classification and speech recognition, where deep learning-based approaches have been extensively studied. One of the main applications is the user-centric platform that consists of great deal of users, items and user generated tabular data which are quite high-dimensional. The characteristics of such data pose unique challenges to the adoption of deep learning in these applications, including modeling, training, and online serving, etc. More and more communities from both academia and industry have initiated the endeavors to solve these challenges. This workshop will provide a venue for both the research and engineering communities to discuss and formulate the challenges, utilize opportunities, and propose new ideas in the practice and theory of deep learning on high-dimensional, sparse and imbalanced data.

In recent years, machine learning methods have been widely used in modern electronic health record (EHR) systems, and have shown more accurate prediction performance on disease risk assessment tasks than traditional methods. However, most of the existing machine learning methods make the assessment solely based on features of the target case but ignore the cross-sample feature interactions between the target case and other similar cases, which is inconsistent with the general practice of evidence-based medicine of making diagnoses based on existing clinical experience. Moreover, current methods that focus on mining cross-sample information rely on deep neural networks to extract cross-sample feature interactions, which would suffer from the problems of data insufficiency, data heterogeneity and lack of interpretability in disease risk assessment tasks. In this work, we propose a novel retrieval-based gradient boosting decision trees (RB-GBDT) model with a cross-sample extractor to mine cross-sample information while exploiting the superiority of GBDT of robustness, generalization and interpretability. Experiments on real-world clinical datasets show the superiority and efficacy of RB-GBDT on disease risk assessment tasks. The developed software has been deployed in hospital as an auxiliary diagnosis tool for risk assessment of venous thromboembolism.

As one of the fundamental trends for future development of recommender systems, Fashion Clothes Matching Recommendation for click-through rate (CTR) prediction has become an increasingly essential task. Unlike traditional single-item recommendation, a combo item, composed of a top item (e.g. a shirt) and a bottom item (e.g. a skirt), is recommended. In such a task, the matching effect between these two single items plays a crucial role, and greatly influences the users' preferences; however, it is usually neglected by previous approaches in CTR prediction. In this work, we tackle this problem by designing a novel algorithm called Combo-Fashion, which extracts the matching effect by introducing the matching history of the combo item with two cascaded modules: (i) Matching Search Module (MSM) seeks the popular combo items and undesirable ones as a positive set and a negative set, respectively; (ii) Matching Prediction Module (MPM) models the precise relationship between the candidate combo item and the positive/negative set by an attention-based deep model. Besides, the CPM Fashion Attribute, considered from characteristic, pattern and material, is applied to capture the matching effect further. As part of this work, we release two large-scale datasets consisting of 3.56 million and 6.01 million user behaviors with rich context and fashion information in millions of combo items. The experimental results over these two real-world datasets have demonstrated the superiority of our proposed model with significant improvements. Furthermore, we have deployed Combo-Fashion onto the platform of Taobao to recommend the combo items to the users, where an 8-day online A/B test proved the effectiveness of Combo-Fashion with an improvement of pCTR by 1.02% and uCTR by 0.70%.

User-tag profile modeling has become one of the novel and significant trends for the future development of industrial recommendation systems, which can be divided into two fundamental tasks: User Preferred Tag (UPT) and Tag Preferred User (TPU) in practical scenarios. In most existing deep learning models for user-tag profiling, the network inputs all the combined tags of the item with the user features when training but inputs only one tag with the user feature to evaluate the user's preference on a single tag when testing. This leads to data discrepancy between the training and testing samples. To address such an issue, we attempt a novel Random Masking Model (RMM) to remain only one tag at the training time by masking. However, it causes two other serious downsides. First, not all tags attached to the same item are equally predictive. Irrelevant tags may introduce noisy signals and thus cause performance degradation. Second, it neglects the impact of combined tags aggregated together, which may be an essential factor leading to user clicks. Therefore, we further propose a framework called Contrast Weighted Tag Masking (CWTM) in this work, which tackles these two issues with two modules: (i) Weighted Masking Module (WMM) introduces the importance network to compute a score for each tag attached to the item and then samples from these tags weightedly according to the score; (ii) Contrast Module (CM) makes use of a contrastive learning architecture to inherit and distill some understanding about the effect of aggregated tags. Offline experiments on four datasets (three public datasets and one proprietary industrial dataset) demonstrate the superiority and effectiveness of CWTM over the state-of-the-art baselines. Moreover, CWTM has been deployed on the training platform of Alibaba advertising systems and achieved substantial improvements of ROI and CVR by 16.8% and 9.6%, respectively.

This paper introduces Honor of Kings Arena, a reinforcement learning (RL) environment based on the Honor of Kings, one of the world’s most popular games at present. Compared to other environments studied in most previous work, ours presents new generalization challenges for competitive reinforcement learning. It is a multi-agent problem with one agent competing against its opponent; and it requires the generalization ability as it has diverse targets to control and diverse opponents to compete with. We describe the observation, action, and reward specifications for the Honor of Kings domain and provide an open-source Python-based interface for communicating with the game engine. We provide twenty target heroes with a variety of tasks in Honor of Kings Arena and present initial baseline results for RL-based methods with feasible computing resources. Finally, we showcase the generalization challenges imposed by Honor of Kings Arena and possible remedies to the challenges. All of the software, including the environment-class, are publicly available.

Prediction over tabular data is an essential and fundamental problem in many important downstream tasks. However, existing methods either take a data instance of the table independently as input or do not fully utilize the multi-row features and labels to directly change and enhance the target data representations. In this paper, we propose to 1) construct a hypergraph from relevant data instance retrieval to model the cross-row and cross-column patterns of those instances, and 2) perform message Propagation to Enhance the target data instance representation for Tabular prediction tasks. Specifically, our specially-designed message propagation step benefits from 1) the fusion of label and features during propagation, and 2) locality-aware multiplicative high-order interaction between features. Experiments on two important tabular prediction tasks validate the superiority of the proposed PET model against other baselines. Additionally, we demonstrate the effectiveness of the model components and the feature enhancement ability of PET via various ablation studies and visualizations. The code is available at https://github.com/KounianhuaDu/PET.

A good state representation is crucial to solving complicated reinforcement learning (RL) challenges. Many recent works focus on designing auxiliary losses for learning informative representations. Unfortunately, these handcrafted objectives rely heavily on expert knowledge and may be sub-optimal. In this paper, we propose a principled and universal method for learning better representations with auxiliary loss functions, named Automated Auxiliary Loss Search (A2LS), which automatically searches for top-performing auxiliary loss functions for RL. Specifically, based on the collected trajectory data, we define a general auxiliary loss space of size $7.5 \times 10^{20}$ and explore the space with an efficient evolutionary search strategy. Empirical results show that the discovered auxiliary loss (namely, A2-winner) significantly improves the performance on both high-dimensional (image) and low-dimensional (vector) unseen tasks with much higher efficiency, showing promising generalization ability to different settings and even different benchmark domains. We conduct a statistical analysis to reveal the relations between patterns of auxiliary losses and RL performance.

Offline reinforcement learning (RL) aims at learning effective policies from historical data without extra environment interactions. During our experience of applying offline RL, we noticed that previous offline RL benchmarks commonly involve significant reality gaps, which we have identified include rich and overly exploratory datasets, degraded baseline, and missing policy validation. In many real-world situations, to ensure system safety, running an overly exploratory policy to collect various data is prohibited, thus only a narrow data distribution is available. The resulting policy is regarded as effective if it is better than the working behavior policy; the policy model can be deployed only if it has been well validated, rather than accomplished the training. In this paper, we present a Near real-world offline RL benchmark, named NeoRL, to reflect these properties. NeoRL datasets are collected with a more conservative strategy. Moreover, NeoRL contains the offline training and offline validation pipeline before the online test, corresponding to real-world situations. We then evaluate recent state-of-the-art offline RL algorithms in NeoRL. The empirical results demonstrate that some offline RL algorithms are less competitive to the behavior cloning and the deterministic behavior policy, implying that they could be less effective in real-world tasks than in the previous benchmarks. We also disclose that current offline policy evaluation methods could hardly select the best policy. We hope this work will shed some light on future research and deploying RL in real-world systems.

Offline reinforcement learning (RL) aims at learning policies from previously collected static trajectory data without interacting with the real environment. Recent works provide a novel perspective by viewing offline RL as a generic sequence generation problem, adopting sequence models such as Transformer architecture to model distributions over trajectories and repurposing beam search as a planning algorithm. However, the training datasets utilized in general offline RL tasks are quite limited and often suffering from insufficient distribution coverage, which could me harmful to training sequence generation models yet has not drawn enough attention in the previous works. In this paper, we propose a novel algorithm named Bootstrapped Transformer, which incorporates the idea of bootstrapping and leverages the learned model to self-generate more offline data to further boost the training of sequence model. We conduct extensive experiments on two offline RL benchmarks and demonstrate that our model can largely remedy the limitations of the existing offline RL training and beat other strong baseline methods. We also analyze the generated pseudo data and the revealed characteristics may shed some light on offline RL training.

Large sequence models (SM) such as GPT series and BERT have displayed outstanding performance and generalization capabilities in natural language process, vision and recently reinforcement learning. A natural follow-up question is how to abstract multi-agent decision making also as an sequence modeling problem and benefit from the prosperous development of the SMs. In this paper, we introduce a novel architecture named Multi-Agent Transformer (MAT) that effectively casts cooperative multi-agent reinforcement learning (MARL) into SM problems wherein the objective is to map agents' observation sequences to agents' optimal action sequences. Our goal is to build the bridge between MARL and SMs so that the modeling power of modern sequence models can be unleashed for MARL. Central to our MAT is an encoder-decoder architecture which leverages the multi-agent advantage decomposition theorem to transform the joint policy search problem into a sequential decision making process; this renders only linear time complexity for multi-agent problems and, most importantly, endows MAT with monotonic performance improvement guarantee. Unlike prior arts such as Decision Transformer fit only pre-collected offline data, MAT is trained by online trial and error from the environment in an on-policy fashion. To validate MAT, we conduct extensive experiments on StarCraftII, Multi-Agent MuJoCo, Dexterous Hands Manipulation, and Google Research Football benchmarks. Results demonstrate that MAT achieves superior performance and data efficiency compared to strong baselines including MAPPO and HAPPO. Furthermore, we demonstrate that MAT is an excellent few-short learner on unseen tasks regardless of changes in the number of agents.See our project page at https://sites.google.com/view/multi-agent-transformer.

As a challenging multi-player card game, DouDizhu has recently drawn much attention for analyzing competition and collaboration in imperfect-information games. In this paper, we propose PerfectDou, a state-of-the-art Doudizhu AI system that summits the game, in an actor-critic framework with a proposed technique named perfect information distillation.In detail, we adopt a perfect-training-imperfection-execution framework that allows the agents to utilize the global information to guide the training of the policies as if it is a perfect information game and the trained policies can be used to play the imperfect information game during the actual gameplay. Correspondingly, we characterize card and game features for DouDizhu to represent the perfect and imperfect information. To train our system, we adopt proximal policy optimization with generalized advantage estimation in a parallel training paradigm. In experiments we show how and why PerfectDou beats all existing programs, and achieves state-of-the-art performance.

With the success of deep learning, click-through rate (CTR) predictions are transitioning from shallow approaches to deep architectures. Current deep CTR prediction usually follows the Embedding & MLP paradigm, where the model embeds categorical features into latent semantic space. This paper introduces a novel embedding technique called neural statistics that instead learns explicit semantics of categorical features by incorporating feature engineering as an innate prior into the deep architecture in an end-to-end manner. Besides, since the statistical information changes over time, we study how to adapt to the distribution shift in the MLP module efficiently. Offline experiments on two public datasets validate the effectiveness of neural statistics against state-of-the-art models. We also apply it to a large-scale recommender system via online A/B tests, where the user's satisfaction is significantly improved.

Building a multi-stage cascade ranking system is a commonly used solution to balance the efficiency and effectiveness in modern information retrieval (IR) applications, such as recommendation and web search. Despite the popularity in practice, the literature specific on multi-stage cascade ranking systems is relatively scarce. The common practice is to train rankers of each stage independently using the same user feedback data (a.k.a., impression data), disregarding the data flow and the possible interactions between stages. This straightforward solution could lead to a sub-optimal system because of the sample selection bias (SSB) issue, which is especially damaging for cascade rankers due to the negative effect accumulated in the multiple stages. Worse still, the interactions between the rankers of each stage are not fully exploited. This paper provides an elaborate analysis of this commonly used solution to reveal its limitations. By studying the essence of cascade ranking, we propose a joint training framework named RankFlow to alleviate the SSB issue and exploit the interactions between the cascade rankers, which is the first systematic solution for this topic. We propose a paradigm of training cascade rankers that emphasizes the importance of fitting rankers on stage-specific data distributions instead of the unified user feedback distribution. We design the RankFlow framework based on this paradigm: The training data of each stage is generated by its preceding stages while the guidance signals not only come from the logs but its successors. Extensive experiments are conducted on various IR scenarios, including recommendation, web search and advertisement. The results verify the efficacy and superiority of RankFlow.

As the final stage of the multi-stage recommender system (MRS), reranking directly affects users' experience and satisfaction, thus playing a critical role in MRS. Despite the improvement achieved in the existing work, three issues are yet to be solved. First, users' historical behaviors contain rich preference information, such as users' long and short-term interests, but are not fully exploited in reranking. Previous work typically treats items in history equally important, neglecting the dynamic interaction between the history and candidate items. Second, existing reranking models focus on learning interactions at the item level while ignoring the fine-grained feature-level interactions. Lastly, estimating the reranking score on the ordered initial list before reranking may lead to the early scoring problem, thereby yielding suboptimal reranking performance. To address the above issues, we propose a framework named Multi-level Interaction Reranking (MIR). MIR combines low-level cross-item interaction and high-level set-to-list interaction, where we view the candidate items to be reranked as a set and the users' behavior history in chronological order as a list. We design a novel SLAttention structure for modeling the set-to-list interactions with personalized long-short term interests. Moreover, feature-level interactions are incorporated to capture the fine-grained influence among items. We design MIR in such a way that any permutation of the input items would not change the output ranking, and we theoretically prove it. Extensive experiments on three public and proprietary datasets show that MIR significantly outperforms the state-of-the-art models using various ranking and utility metrics.

Information retrieval (IR) systems have become an essential component in modern society to help users find useful information, which consists of a series of processes including query expansion, item recall, item ranking and re-ranking, etc. Based on the ranked information list, users can provide their feedbacks. Such an interaction process between users and IR systems can be naturally formulated as a decision-making problem, which can be either one-step or sequential. In the last ten years, deep reinforcement learning (DRL) has become a promising direction for decision-making, since DRL utilizes the high model capacity of deep learning for complex decision-making tasks. Recently, there have been emerging research works focusing on leveraging DRL for IR tasks. However, the fundamental information theory under DRL settings, the principle of RL methods for IR tasks, or the experimental evaluation protocols of DRL-based IR systems, has not been deeply investigated.

Knowledge tracing, which estimates students' knowledge states by predicting the probability that they correctly answer questions, is an essential task for online learning platforms. It has gained much attention in the decades due to its importance to downstream tasks like learning material arrangement, etc. The previous deep learning-based methods trace students' knowledge states with the explicitly intra-student information, i.e., they only consider the historical information of individuals to make predictions. However, they neglect the inter-student information, which contains the response correctness of other students who have similar question-answering experiences, may offer some valuable clues. Based on this consideration, we propose a method called Collaborative Knowledge Tracing (CoKT) in this paper, which sufficiently exploits the inter-student information in knowledge tracing. It retrieves the sequences of peer students who have similar question-answering experiences to obtain the inter-student information, and integrates the inter-student information with the intra-student information to trace students' knowledge states and predict their correctness in answering questions. We validate the effectiveness of our method on four real-world datasets and compare it with 11 baselines. The experimental results reveal that CoKT achieves the best performance.

The personalized recommendation is an essential part of modern e-commerce, where user’s demands are not only conditioned by their profile but also by their recent browsing behaviors as well as periodical purchases made some time ago. In this paper, we propose a novel framework named Search-based Time-Aware Recommendation (STARec), which captures the evolving demands of users over time through a unified search-based time-aware model. More concretely, we first design a search-based module to retrieve a user’s relevant historical behaviors, which are then mixed up with her recent records to be fed into a time-aware sequential network for capturing her time-sensitive demands. Besides retrieving relevant information from her personal history, we also propose to search and retrieve similar user’s records as an additional reference. All these sequential records are further fused to make the final recommendation. Beyond this framework, we also develop a novel label trick that uses the previous labels (i.e., user’s feedbacks) as the input to better capture the user’s browsing pattern. We conduct extensive experiments on three real-world commercial datasets on click-through-rate prediction tasks against state-of-the-art methods. Experimental results demonstrate the superiority and efficiency of our proposed framework and techniques. Furthermore, results of online experiments on a daily item recommendation platform of Company X show that STARec gains average performance improvement of around 6% and 1.5% in its two main item recommendation scenarios on CTR metric respectively.

With the prevalence of live broadcast business nowadays, a new type of recommendation service, called live broadcast recommendation, is widely used in many mobile e-commerce Apps. Different from classical item recommendation, live broadcast recommendation is to automatically recommend user anchors instead of items considering the interactions among triple-objects (i.e., users, anchors, items) rather than binary interactions between users and items. Existing methods based on binary objects, ranging from early matrix factorization to recently emerged deep learning, obtain objects’ embeddings by mapping from pre-existing features. Directly applying these techniques would lead to limited performance, as they are failing to encode collaborative signals among triple-objects. In this paper, we propose a novel TWo-side Interactive NetworkS (TWINS) for live broadcast recommendation. In order to fully use both static and dynamic information on user and anchor sides, we combine a product-based neural network with a recurrent neural network to learn the embedding of each object. In addition, instead of directly measuring the similarity, TWINS effectively injects the collaborative effects into the embedding process in an explicit manner by modeling interactive patterns between the user’s browsing history and the anchor’s broadcast history in both item and anchor aspects. Furthermore, we design a novel co-retrieval technique to select key items among massive historic records efficiently. Offline experiments on real large-scale data show the superior performance of the proposed TWINS, compared to representative methods; and further results of online experiments on Diantao App show that TWINS gains average performance improvement of around 8% on ACTR metric, 3% on UCTR metric, 3.5% on UCVR metric.

Venous thromboembolism (VTE) is the world's third most common cause of vascular mortality and a serious complication from multiple departments. Risk assessment of VTE guides clinical intervention in time and is of great importance to in-hospital patients. Traditional VTE risk assessment methods based on scaling tools, which always require rules carefully designed by human experts, are difficult to apply to large-population scenarios since the manually designed rules are not guaranteed to be accurate to all populations. In contrast, with the development of the electronic health record (EHR) datasets, data-driven machine-learning-based risk assessment methods have proven superior predictability in many studies in recent years. This paper uses the gradient boosting tree model to study the VTE risk assessment problem with multi-department data. There exist two distinct characteristics of VTE data collected at the level of the entire hospital: its wide distribution and heterogeneity across multiple departments. To this end, we consider the prediction task over multiple departments as a multi-task learning process, and introduce the algorithm of a task-aware tree-based method TSGB to tackle the multi-task prediction problem. Although the introduction of multi-task learning improves overall across-department performance, we reveal the problem of task-wise performance decline while dealing with imbalanced VTE data volume. According to the analysis, we finally propose two variants of TSGB to alleviate the problems and further boost the prediction performance. Compared with state-of-the-art rule-based and multi-task tree-based methods, the experimental results show the proposed methods not only improve the overall across-department AUC performance effectively, but also ensure the improvement of performance over every single department prediction.

Predicting urban traffic is of great importance to intelligent transportation systems and public safety, yet is very challenging in three aspects: 1) complex spatio-temporal correlations of urban traffic, including spatial correlations between locations along with temporal correlations among timestamps; 2) spatial diversity of such spatio-temporal correlations, which varies from location to location and depends on the surrounding geographical information, e.g., points of interests and road networks; and 3) temporal diversity of such spatio-temporal correlations, which is highly influenced by dynamic traffic states. To tackle these challenges, we proposed a deep meta learning based model, entitled ST-MetaNet$^+$+, tocollectivelypredict traffic in all locations at the same time. ST-MetaNet$^+$+employs a sequence-to-sequence architecture, consisting of an encoder to learn historical information and a decoder to make predictions step by step. Specifically, the encoder and decoder have the same network structure, consisting of meta graph attention networks and meta recurrent neural networks, to capture diverse spatial and temporal correlations, respectively. Furthermore, the weights (parameters) of meta graph attention networks and meta recurrent neural networks are generated from the embeddings of geo-graph attributes and the traffic context learned from dynamic traffic states. Extensive experiments were conducted based on three real-world datasets to illustrate the effectiveness of ST-MetaNet$^+$+beyond several state-of-the-art methods.

In this article, we study the problem of guaranteed display ads (GDAs) allocation, which requires proactively allocate display ads to different impressions to fulfill their impression demands indicated in the contracts. Existing methods for this problem either assume the impressions that are static or solely consider a specific ad's benefits. Thus, it is hard to generalize to the industrial production scenario where the impressions are dynamical and large-scale, and the overall allocation optimality of all the considered GDAs is required. To bridge this gap, we formulate this problem as a sequential decision-making problem in the scope of multiagent reinforcement learning (MARL), by assigning an allocation agent to each ad and coordinating all the agents for allocating GDAs. The inputs are the states (e.g., the demands of the ad and the remaining time steps for displaying the ads) of each ad and the impressions at different time steps, and the outputs are the display ratios of each ad for each impression. Specifically, we propose a novel hierarchical MARL (HMARL) method that creates hierarchies over the agent policies to handle a large number of ads and the dynamics of impressions. HMARL contains: 1) a manager policy to navigate the agent to choose an appropriate subpolicy and 2) a set of subpolicies that let the agents perform diverse conditioning on their states. Extensive experiments on three real-world data sets from the Tencent advertising platform with tens of millions of records demonstrate significant improvements of HMARL over state-of-the-art approaches.

Learning to rank from logged user feedback, such as clicks or purchases, is a central component of many real-world information systems. Different from human-annotated relevance labels, the user feedback is always noisy and biased. Many existing learning to rank methods infer the underlying relevance of query–item pairs based on different assumptions of examination, and still optimize a relevance based objective. Such methods rely heavily on the correct estimation of examination, which is often difficult to achieve in practice. In this work, we propose a general framework U-rank+ for learning to rank with logged user feedback from the perspective of graph matching. We systematically analyze the biases in user feedback, including examination bias and selection bias. Then, we take both biases into consideration for unbiased utility estimation that directly based on user feedback, instead of relevance. In order to maximize the estimated utility in an efficient manner, we design two different solvers based on Sinkhorn and LambdaLoss for U-rank+ . The former is based on a standard graph matching algorithm, and the latter is inspired by the traditional method of learning to rank. Both of the algorithms have good theoretical properties to optimize the unbiased utility objective while the latter is proved to be empirically more effective and efficient in practice. Our framework U-rank+ can deal with a general utility function and can be used in a widespread of applications including web search, recommendation, and online advertising. Semi-synthetic experiments on three benchmark learning to rank datasets demonstrate the effectiveness of U-rank+ . Furthermore, our proposed framework has been deployed on two different scenarios of a mainstream App store, where the online A/B testing shows that U-rank+ achieves an average improvement of 19.2% on click-through rate and 20.8% improvement on conversion rate in recommendation scenario, and 5.12% on platform revenue in online advertising scenario over the production baselines.

Heterogeneous information network (HIN) has been widely used to characterize entities of various types and their complex relations. Recent attempts either rely on explicit path reachability to leverage path-based semantic relatedness or graph neighborhood to learn heterogeneous network representations before predictions. These weakly coupled manners overlook the rich interactions among neighbor nodes, which introduces an early summarization issue. In this article, we propose GraphHINGE ( H eterogeneous IN teract and aggre G at E ), which captures and aggregates the interactive patterns between each pair of nodes through their structured neighborhoods. Specifically, we first introduce Neighborhood-based Interaction (NI) module to model the interactive patterns under the same metapaths, and then extend it to Cross Neighborhood-based Interaction (CNI) module to deal with different metapaths. Next, in order to address the complexity issue on large-scale networks, we formulate the interaction modules via a convolutional framework and learn the parameters efficiently with fast Fourier transform. Furthermore, we design a novel neighborhood-based selection (NS) mechanism, a sampling strategy, to filter high-order neighborhood information based on their low-order performance. The extensive experiments on six different types of heterogeneous graphs demonstrate the performance gains by comparing with state-of-the-arts in both click-through rate prediction and top-N recommendation tasks.

Wildlife poaching is threatening key species that play important roles in the ecosystem. With historical ranger patrol records, it is possible to provide data-driven predictions of poaching threats and plan patrols to combat poaching. However, the patrollers often patrol in a multimodal way, which combines driving and walking. It is a tedious task for the domain experts to manually plan such a patrol and as a result, the planned patrol routes are often far from optimal. In this paper, we propose a data-driven approach for multimodal patrol planning. We first use machine learning models to predict the poaching threats and then use a novel mixed-integer linear programming-based algorithm to plan the patrol route. In a field test focusing on the machine learning prediction result at Jilin Huangnihe National Nature Reserve (HNHR) in December 2019, the rangers found 42 snares, which is significantly higher than the historical record. Our offline experiments show that the resulting multimodal patrol routes can improve the efficiency of patrol and thus they can serve as the basis for future deployment in the field.

As a fundamental problem in algorithmic trading, order execution aims at fulfilling a specific trading order, either liquidation or acquirement, for a given instrument. Towards effective execution strategy, recent years have witnessed the shift from the analytical view with model-based market assumptions to model-free perspective, i.e., reinforcement learning, due to its nature of sequential decision optimization. However, the noisy and yet imperfect market information that can be leveraged by the policy has made it quite challenging to build up sample efficient reinforcement learning methods to achieve effective order execution. In this paper, we propose a novel universal trading policy optimization framework to bridge the gap between the noisy yet imperfect market states and the optimal action sequences for order execution. Particularly, this framework leverages a policy distillation method that can better guide the learning of the common policy towards practically optimal execution by an oracle teacher with perfect information to approximate the optimal trading strategy. The extensive experiments have shown significant improvements of our method over various strong baselines, with reasonable trading actions.

Lihua Qian, Hao Zhou, Yu Bao, Mingxuan Wang, Lin Qiu, Weinan Zhang, Yong Yu, Lei Li. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Cycle-consistent training is widely used for jointly learning a forward and inverse mapping between two domains of interest without the cumbersome requirement of collecting matched pairs within each domain. In this regard, the implicit assumption is that there exists (at least approximately) a ground-truth bijection such that a given input from either domain can be accurately reconstructed from successive application of the respective mappings. But in many applications no such bijection can be expected to exist and large reconstruction errors can compromise the success of cycle-consistent training. As one important instance of this limitation, we consider practically-relevant situations where there exists a many-to-one or surjective mapping between domains. To address this regime, we develop a conditional variational autoencoder (CVAE) approach that can be viewed as converting surjective mappings to implicit bijections whereby reconstruction errors in both directions can be minimized, and as a natural byproduct, realistic output diversity can be obtained in the one-to-many direction. As theoretical motivation, we analyze a simplified scenario whereby minima of the proposed CVAE-based energy function align with the recovery of ground-truth surjective mappings. On the empirical side, we consider a synthetic image dataset with known ground-truth, as well as a real-world application involving natural language generation from knowledge graphs and vice versa, a prototypical surjective case. For the latter, our CVAE pipeline can capture such many-to-one mappings during cycle training while promoting textural diversity for graph-to-text tasks.

The research of geoscience plays a strong role in helping people gain a better understanding of the Earth. To effectively represent the knowledge (KG) from enormous geoscience research papers, knowledge graphs can be a powerful means. In the face of enormous geoscience research papers, knowledge graphs can be a powerful means to manage the relationships of data and integrate knowledge extracted from them. However, the existing geoscience KGs mainly focus on the external connection between concepts, whereas the potential abundant information contained in the internal multimodal data of the paper is largely overlooked for more fine-grained knowledge mining. To this end, we propose GAKG, a large-scale multimodal academic KG based on 1.12 million papers published in various geoscience-related journals. In addition to the bibliometrics elements, we also extracted the internal illustrations, tables, and text information of the articles, and dig out the knowledge entities of the papers and the era and spatial attributes of the articles, coupling multimodal academic data and features. Specifically, GAKG realizes knowledge entity extraction under our proposed Human-In-the-Loop framework, the novelty of which is to combine the techniques of machine reading and information retrieval with manual annotation of geoscientists in the loop. Considering the fact that literature of geoscience often contains more abundant illustrations and time scale information compared with that of other disciplines, we extract all the geographical information and era from the geoscience papers' text and illustrations, mapping papers to the atlas and chronology. Based on GAKG, we build several knowledge discovery benchmarks for finding geoscience communities and predicting potential links. GAKG and its services have been made publicly available and user-friendly.

In this paper, we introduce LiteratureQA, a large question answering (QA) corpus consisting of publicly available academic papers. Different from other QA corpus, LiteratureQA has its unique challenges such as how to leverage the structured knowledge of citation networks. We further examine some popular QA method and present a benchmark approach of answering academic questions by combining both semantic text and graph knowledge to improve the prevalent pre-training model. We hope this resource could help research and development of tasks for machine reading over academic text.

Document-level relation extraction aims to identify relations between entities in a whole document.Prior efforts to capture long-range dependencies have relied heavily on implicitly powerful representations learned through (graph) neural networks, which makes the model less transparent.To tackle this challenge, in this paper, we propose LogiRE, a novel probabilistic model for document-level relation extraction by learning logic rules.Lo-giRE treats logic rules as latent variables and consists of two modules: a rule generator and a relation extractor.The rule generator is to generate logic rules potentially contributing to final predictions, and the relation extractor outputs final predictions based on the generated logic rules.Those two modules can be efficiently optimized with the expectationmaximization (EM) algorithm.By introducing logic rules into neural networks, LogiRE can explicitly capture long-range dependencies as well as enjoy better interpretation.Empirical results show that LogiRE significantly outperforms several strong baselines in terms of relation performance (∼1.8 F1 score) and logical consistency (over 3.3 logic score).Our code is available at https://github.com/rudongyu/LogiRE.

This paper investigates the model-based methods in multi-agent reinforcement learning (MARL). We specify the dynamics sample complexity and the opponent sample complexity in MARL, and conduct a theoretic analysis of return discrepancy upper bound. To reduce the upper bound with the intention of low sample complexity during the whole learning process, we propose a novel decentralized model-based MARL method, named Adaptive Opponent-wise Rollout Policy Optimization (AORPO). In AORPO, each agent builds its multi-agent environment model, consisting of a dynamics model and multiple opponent models, and trains its policy with the adaptive opponent-wise rollout. We further prove the theoretic convergence of AORPO under reasonable assumptions. Empirical experiments on competitive and cooperative tasks demonstrate that AORPO can achieve improved sample efficiency with comparable asymptotic performance over the compared MARL methods.

Click-through rate (CTR) estimation plays as a core function module in various personalized online services, including online advertising, recommender systems, and web search etc. From 2015, the success of deep learning started to benefit CTR estimation performance and now deep CTR models have been widely applied in many industrial platforms. In this survey, we provide a comprehensive review of deep learning models for CTR estimation tasks. First, we take a review of the transfer from shallow to deep CTR models and explain why going deep is a necessary trend of development. Second, we concentrate on explicit feature interaction learning modules of deep CTR models. Then, as an important perspective on large platforms with abundant user histories, deep behavior models are discussed. Moreover, the recently emerged automated methods for deep CTR architecture design are presented. Finally, we summarize the survey and discuss the future prospects of this field.

In Goal-oriented Reinforcement learning, relabeling the raw goals in past experience to provide agents with hindsight ability is a major solution to the reward sparsity problem. In this paper, to enhance the diversity of relabeled goals, we develop FGI (Foresight Goal Inference), a new relabeling strategy that relabels the goals by looking into the future with a learned dynamics model. Besides, to improve sample efficiency, we propose to use the dynamics model to generate simulated trajectories for policy training. By integrating these two improvements, we introduce the MapGo framework (Model-Assisted Policy optimization for Goal-oriented tasks). In our experiments, we first show the effectiveness of the FGI strategy compared with the hindsight one, and then show that the MapGo framework achieves higher sample efficiency when compared to model-free baselines on a set of complicated tasks.

Diabetes prediction is an important data science application in the social healthcare domain. There exist two main challenges in the diabetes prediction task: data heterogeneity since demographic and metabolic data are of different types, data insufficiency since the number of diabetes cases in a single medical center is usually limited. To tackle the above challenges, we employ gradient boosting decision trees (GBDT) to handle data heterogeneity and introduce multi-task learning (MTL) to solve data insufficiency. To this end, Task-wise Split Gradient Boosting Trees (TSGB) is proposed for the multi-center diabetes prediction task. Specifically, we firstly introduce task gain to evaluate each task separately during tree construction, with a theoretical analysis of GBDT's learning objective. Secondly, we reveal a problem when directly applying GBDT in MTL, i.e., the negative task gain problem. Finally, we propose a novel split method for GBDT in MTL based on the task gain statistics, named task-wise split, as an alternative to standard feature-wise split to overcome the mentioned negative task gain problem. Extensive experiments on a large-scale real-world diabetes dataset and a commonly used benchmark dataset demonstrate TSGB achieves superior performance against several state-of-the-art methods. Detailed case studies further support our analysis of negative task gain problems and provide insightful findings. The proposed TSGB method has been deployed as an online diabetes risk assessment software for early diagnosis.

Click-Through Rate (CTR) prediction is critical for industrial recommender systems, where most deep CTR models follow an Embedding & Feature Interaction paradigm. However, the majority of methods focus on designing network architectures to better capture feature interactions while the feature embedding, especially for numerical features, has been overlooked. Existing approaches for numerical features are difficult to capture informative knowledge because of the low capacity or hard discretization based on the offline expertise feature engineering. In this paper, we propose a novel embedding learning framework for numerical features in CTR prediction (AutoDis) with high model capacity, end-to-end training and unique representation properties preserved. AutoDis consists of three core components: meta-embeddings, automatic discretization and aggregation. Specifically, we propose meta-embeddings for each numerical field to learn global knowledge from the perspective of field with a manageable number of parameters. Then the differentiable automatic discretization performs soft discretization and captures the correlations between the numerical features and meta-embeddings. Finally, distinctive and informative embeddings are learned via an aggregation function. Comprehensive experiments on two public and one industrial datasets are conducted to validate the effectiveness of AutoDis. Moreover, AutoDis has been deployed onto a mainstream advertising platform, where online A/B test demonstrates the improvement over the base model by 2.1% and 2.7% in terms of CTR and eCPM, respectively. In addition, the code of our framework is publicly available in MindSpore.

Prediction over tabular data is an essential task in many data science applications such as recommender systems, online advertising, medical treatment, etc. Tabular data is structured into rows and columns, with each row as a data sample and each column as a feature attribute. Both the columns and rows of the tabular data carry useful patterns that could improve the model prediction performance. However, most existing models focus on the cross-column patterns yet overlook the cross-rowpatterns as they deal with single samples independently. In this work, we propose a general learning framework named Retrieval & Interaction Machine (RIM) that fully exploits both cross-row and cross-column patterns among tabular data. Specifically, RIM first leverages search engine techniques to efficiently retrieve useful rows of the table to assist the label prediction of the target row, then uses feature interaction networks to capture the cross-column patterns among the target row and the retrieved rows so as to make the final label prediction. We conduct extensive experiments on 11 datasets of three important tasks, i.e., CTR prediction (classification), top-n recommendation (ranking) and rating prediction (regression). Experimental results show that RIM achieves significant improvements over the state-of-the-art and various baselines, demonstrating the superiority and efficacy of RIM.

Recently, we have witnessed that deep learning-based approaches has been widely applied to empower many internet-scale applications. However, the data in these internet-scale applications are high dimensional and extremely sparse, which makes it different from those applications with dense data processing, such as image classification and speech recognition, where deep learning-based approaches have been extensively studied. One of the main applications is the user-centric platform that consists of great deal of users, items and user generated tabular data which are quite high-dimensional. The characteristics of such data pose unique challenges to the adoption of deep learning in these applications, including modeling, training, and online serving, etc. More and more communities from both academia and industry have initiated the endeavors to solve these challenges. This workshop will provide a venue for both the research and engineering communities to discuss and formulate the challenges, utilize opportunities, and propose new ideas in the practice of deep learning on high-dimensional sparse data.

Model-based reinforcement learning has attracted wide attention due to its superior sample efficiency. Despite its impressive success so far, it is still unclear how to appropriately schedule the important hyperparameters to achieve adequate performance, such as the real data ratio for policy optimization in Dyna-style model-based algorithms. In this paper, we first theoretically analyze the role of real data in policy training, which suggests that gradually increasing the ratio of real data yields better performance. Inspired by the analysis, we propose a framework named AutoMBPO to automatically schedule the real data ratio as well as other hyperparameters in training model-based policy optimization (MBPO) algorithm, a representative running case of model-based methods. On several continuous control tasks, the MBPO instance trained with hyperparameters scheduled by AutoMBPO can significantly surpass the original one, and the real data ratio schedule found by AutoMBPO shows consistency with our theoretical analysis.

Offline reinforcement learning (RL) tasks require the agent to learn from a pre-collected dataset with no further interactions with the environment. Despite the potential to surpass the behavioral policies, RL-based methods are generally impractical due to the training instability and bootstrapping the extrapolation errors, which always require careful hyperparameter tuning via online evaluation. In contrast, offline imitation learning (IL) has no such issues since it learns the policy directly without estimating the value function by bootstrapping. However, IL is usually limited in the capability of the behavioral policy and tends to learn a mediocre behavior from the dataset collected by the mixture of policies. In this paper, we aim to take advantage of IL but mitigate such a drawback. Observing that behavior cloning is able to imitate neighboring policies with less data, we propose \textit{Curriculum Offline Imitation Learning (COIL)}, which utilizes an experience picking strategy to make the agent imitate from adaptive neighboring policies with a higher return, and improves the current policy along curriculum stages. On continuous control benchmarks, we compare COIL against both imitation-based methods and RL-based methods, showing that COIL not only avoids just learning a mediocre behavior on mixed datasets but is also even competitive with state-of-the-art offline RL methods.

Modern information retrieval (IR) consists of a series of processes, including query expansion, candidate item recall, item ranking, item re-ranking, etc. The final ranked item list will be exposed to the user, which will accordingly provide feedback through some expected actions such as browsing and click. Such a whole process can be formulated as a decision-making process where the agent is the IR system while the environment is the specific user. This decision-making process can be one-step or sequential, depending on the scenarios or the ways of problem formulation. Since 2013, Deep reinforcement learning (DRL) has been a fast-developing technique for decision-making tasks. The high capacity of deep learning models is incorporated in the reinforcement learning framework so that the agent may successfully handle complex decision-making. In recent years, there have been a bunch of publications attempting to leverage DRL techniques for different IR tasks such as ad hoc retrieval, learning to rank and interactive recommendation. Nonetheless, the fundamental theory, the principle of RL methods or the recognized experimental protocols of decision-making in IR, has not been well developed, making it challenging to evaluate the correctness of a proposed method or judge whether the reported experimental performance is valid. We propose the second DRL4IR workshop at SIGIR 2021, which provides a venue to gather the academia researchers and industry practitioners to present the recent progress of DRL techniques for IR. More importantly, people in this workshop are expected to discuss more about the fundamental principles of formulating a decision-making IR task, the underlying theory as well as the practical effectiveness of the experiment protocol design, which would foster further research on novel methodologies, innovative experimental findings and new applications of DRL for information retrieval. DRL4IR organized at SIGIR'20 was one of the most popular workshops and attracted over 200 conference attendees. In this year, we will pay more attention to fundamental research topics and recent applications, and expect about 300 participants.

To better exploit search logs and model users' behavior patterns, numerous click models are proposed to extract users' implicit interaction feedback. Most traditional click models are based on the probabilistic graphical model (PGM) framework, which requires manually designed dependencies and may oversimplify user behaviors. Recently, methods based on neural networks are proposed to improve the prediction accuracy of user behaviors by enhancing the expressive ability and allowing flexible dependencies. However, they still suffer from the data sparsity and cold-start problems. In this paper, we propose a novel graph-enhanced click model (GraphCM) for web search. Firstly, we regard each query or document as a vertex, and propose novel homogeneous graph construction methods for queries and documents respectively, to fully exploit both intra-session and inter-session information for the sparsity and cold-start problems. Secondly, following the examination hypothesis, we separately model the attractiveness estimator and examination predictor to output the attractiveness scores and examination probabilities, where graph neural networks and neighbor interaction techniques are applied to extract the auxiliary information encoded in the pre-constructed homogeneous graphs. Finally, we apply combination functions to integrate examination probabilities and attractiveness scores into click predictions. Extensive experiments conducted on three real-world session datasets show that GraphCM not only outperforms the state-of-art models, but also achieves superior performance in addressing the data sparsity and cold-start problems.

Knowledge tracing, which dynamically estimates students' learning states by predicting their performance on answering questions, is an essential task in online education. One typical solution for knowledge tracing is based on Recurrent Neural Networks (RNNs), which represent students' knowledge states with the hidden states of RNNs. Such type of methods normally assumes that students have the same cognition level and knowledge acquisition sensitivity on the same question. Thus, they (i) predict students' responses by referring to their knowledge states and question representations, and (ii) update the knowledge states according to the question representations and students' responses. No explicit cognition level or knowledge acquisition sensitivity is considered in the above two processes. However, in real-world scenarios, students have different understandings on a question and have various knowledge acquisition after they finish the same question. In this paper, we propose a novel model called Individual Estimation Knowledge Tracing (IEKT), which estimates the students' cognition on the question before response prediction and assesses their knowledge acquisition sensitivity on the questions before updating the knowledge state. In the experiments, we compare IEKT with 11 knowledge tracing baselines on four benchmark datasets, and the results show IEKT achieves the state-of-the-art performance.

Modern information retrieval systems, including web search, ads placement, and recommender systems, typically rely on learning from user feedback. Click models, which study how users interact with a ranked list of items, provide a useful understanding of user feedback for learning ranking models. Constructing ”right” dependencies is the key of any successful click model. However, probabilistic graphical models (PGMs) have to rely on manually assigned dependencies, and oversimplify user behaviors. Existing neural network based methods promote PGMs by enhancing the expressive ability and allowing flexible dependencies, but still suffer from exposure bias and inferior estimation. In this paper, we propose a novel framework, Adversarial Imitation Click Model (AICM), based on imitation learning. Firstly, we explicitly learn the reward function that recovers users’ intrinsic utility and underlying intentions. Secondly, we model user interactions with a ranked list as a dynamic system instead of one-step click prediction, alleviating the exposure bias problem. Finally, we minimize the JS divergence through adversarial training and learn a stable distribution of click sequences, which makes AICM generalize well across different distributions of ranked lists. A theoretical analysis has indicated that AICM reduces the exposure bias from O(T2) to O(T). Our studies on a public web search dataset show that AICM not only outperforms state-of-the-art models in traditional click metrics but also achieves superior performance in addressing the exposure bias and recovering the underlying patterns of click sequences.

Hero drafting is essential in multiplayer online battle arena (MOBA) game playing as it builds the team of each side and directly affects the match outcome. State-of-the-art drafting methods fail to consider: 1) drafting efficiency when the hero pool is expanded; 2) the multiround nature of a MOBA 5v5 match series, i.e., two teams play best-of-$N$and the same hero is only allowed to be drafted once throughout the series. In this article, we formulate the drafting process as a multiround combinatorial game and propose a novel drafting algorithm based on neural networks and Monte Carlo tree search, named JueWuDraft. Specifically, we design a long-term value estimation mechanism to handle the best-of-$N$drafting case. TakingHonor of Kings, one of the most popular MOBA games at present, as a running case, we demonstrate the practicality and effectiveness of JueWuDraft when compared to state-of-the-art drafting methods.

Graph representation learning aims to embed each vertex in a graph into a low-dimensional vector space. Existing graph representation learning methods can be classified into two categories: generative models that learn the underlying connectivity distribution in a graph, and discriminative models that predict the probability of edge between a pair of vertices. In this paper, we propose GraphGAN, an innovative graph representation learning framework unifying the above two classes of methods, in which the generative and the discriminative model play a game-theoretical minimax game. Specifically, for a given vertex, the generative model tries to fit its underlying true connectivity distribution over all other vertices and produces “fake” samples to fool the discriminative model, while the discriminative model tries to detect whether the sampled vertex is from ground truth or generated by the generative model. With the competition between these two models, both of them can alternately and iteratively boost their performance. Moreover, we propose a novel graph softmax as the implementation of the generative model to overcome the limitations of traditional softmax function, which can be proven satisfying desirable properties of normalization, graph structure awareness, and computational efficiency. Through extensive experiments on real-world datasets, we demonstrate that GraphGAN achieves substantial gains in a variety of applications, including graph reconstruction, link prediction, node classification, recommendation, and visualization, over state-of-the-art baselines.

Author name ambiguity causes inadequacy and inconvenience in academic information retrieval, which raises the necessity of author name disambiguation (AND). Existing AND methods can be divided into two categories: the models focusing on content information to distinguish whether two papers are written by the same author, the models focusing on relation information to represent information as edges on the network and to quantify the similarity among papers. However, the former requires adequate labeled samples and informative negative samples, and are also ineffective in measuring the high-order connections among papers, while the latter needs complicated feature engineering or supervision to construct the network. We propose a novel generative adversarial framework to grow the two categories of models together: (i) the discriminative module distinguishes whether two papers are from the same author, and (ii) the generative module selects possibly homogeneous papers directly from the heterogeneous information network, which eliminates the complicated feature engineering. In such a way, the discriminative module guides the generative module to select homogeneous papers, and the generative module generates high-quality negative samples to train the discriminative module to make it aware of high-order connections among papers. Furthermore, a self-training strategy for the discriminative module and a random walk based generating algorithm are designed to make the training stable and efficient. Extensive experiments on two real-world AND benchmarks demonstrate that our model provides significant performance improvement over the state-of-the-art methods.

GPT-2 and BERT demonstrate the effectiveness of using pre-trained language models (LMs) on various natural language processing tasks. However, LM fine-tuning often suffers from catastrophic forgetting when applied to resource-rich tasks. In this work, we introduce a concerted training framework (CTnmt) that is the key to integrate the pre-trained LMs to neural machine translation (NMT). Our proposed CTnmt} consists of three techniques: a) asymptotic distillation to ensure that the NMT model can retain the previous pre-trained knowledge; b) a dynamic switching gate to avoid catastrophic forgetting of pre-trained knowledge; and c) a strategy to adjust the learning paces according to a scheduled policy. Our experiments in machine translation show CTnmt gains of up to 3 BLEU score on the WMT14 English-German language pair which even surpasses the previous state-of-the-art pre-training aided NMT by 1.4 BLEU score. While for the large WMT14 English-French task with 40 millions of sentence-pairs, our base model still significantly improves upon the state-of-the-art Transformer big model by more than 1 BLEU score.

Coordination is one of the essential problems in multi-agent systems. Typically multi-agent reinforcement learning (MARL) methods treat agents equally and the goal is to solve the Markov game to an arbitrary Nash equilibrium (NE) when multiple equilibra exist, thus lacking a solution for NE selection. In this paper, we treat agents unequally and consider Stackelberg equilibrium as a potentially better convergence point than Nash equilibrium in terms of Pareto superiority, especially in cooperative environments. Under Markov games, we formally define the bi-level reinforcement learning problem in finding Stackelberg equilibrium. We propose a novel bi-level actor-critic learning method that allows agents to have different knowledge base (thus intelligent), while their actions still can be executed simultaneously and distributedly. The convergence proof is given, while the resulting learning algorithm is tested against the state of the arts. We found that the proposed bi-level actor-critic algorithm successfully converged to the Stackelberg equilibria in matrix games and find a asymmetric solution in a highway merge environment.

The CUR matrix decomposition is an important tool for low-rank matrix approximation. It approximates a data matrix though selecting a small number of columns and rows of the matrix. Those CUR algorithms with gap-dependent approximation bounds can obtain high approximation quality for matrices with good singular value spectrum decay, but they have impractically high time complexities. In this paper, we propose a novel CUR algorithm based on truncated LU factorization with an efficient variant of complete pivoting. Our algorithm has gap-dependent approximation bounds on both spectral and Frobenius norms while maintaining high efficiency. Numerical experiments demonstrate the effectiveness of our algorithm and verify our theoretical guarantees.

Learning to rank with implicit feedback is one of the most important tasks in many real-world information systems where the objective is some specific utility, e.g., clicks and revenue. However, we point out that existing methods based on probabilistic ranking principle do not necessarily achieve the highest utility. To this end, we propose a novel ranking framework called U-rank that directly optimizes the expected utility of the ranking list. With a position-aware deep click-through rate prediction model, we address the attention bias considering both query-level and item-level features. Due to the item-specific attention bias modeling, the optimization for expected utility corresponds to a maximum weight matching on the item-position bipartite graph. We base the optimization of this objective in an efficient Lambdaloss framework, which is supported by both theoretical and empirical analysis. We conduct extensive experiments for both web search and recommender systems over three benchmark datasets and two proprietary datasets, where the performance gain of U-rank over state-of-the-arts is demonstrated. Moreover, our proposed U-rank has been deployed on a large-scale commercial recommender and a large improvement over the production baseline has been observed in an online A/B testing.

To drive purchase in online advertising, it is of the advertiser's great interest to optimize the sequential advertising strategy whose performance and interpretability are both important. The lack of interpretability in existing deep reinforcement learning methods makes it not easy to understand, diagnose and further optimize the strategy.In this paper, we propose our Deep Intents Sequential Advertising (DISA) method to address these issues. The key part of interpretability is to understand a consumer's purchase intent which is, however, unobservable (called hidden states). In this paper, we model this intention as a latent variable and formulate the problem as a Partially Observable Markov Decision Process (POMDP) where the underlying intents are inferred based on the observable behaviors. Large-scale industrial offline and online experiments demonstrate our method's superior performance over several baselines. The inferred hidden states are analyzed, and the results prove the rationality of our inference.

The heavy traffic congestion problem has always been a concern for modern cities. To alleviate traffic congestion, researchers use reinforcement learning (RL) to develop better traffic signal control (TSC) algorithms in recent years. However, most RL models are trained and tested in the same traffic flow environment, which results in a serious overfitting problem. Since the traffic flow environment in the real world keeps varying, these models can hardly be applied due to the lack of generalization ability. Besides, the limited number of accessible traffic flow data brings extra difficulty in testing the generalization ability of the models. In this paper, we design a novel traffic flow generator based on Wasserstein generative adversarial network to generate sufficient diverse and quality traffic flows and use them to build proper training and testing environments. Then we propose a meta-RL TSC framework GeneraLight to improve the generalization ability of TSC models. GeneraLight boosts the generalization performance by combining the idea of flow clustering and model-agnostic meta-learning. We conduct extensive experiments on multiple real-world datasets to show the superior performance of GeneraLight on generalizing to different traffic flows.

Domain adaptation aims to leverage the supervision signal of source domain to obtain an accurate model for target domain, where the labels are not available. To leverage and adapt the label information from source domain, most existing methods employ a feature extracting function and match the marginal distributions of source and target domains in a shared feature space. In this paper, from the perspective of information theory, we show that representation matching is actually an insufficient constraint on the feature space for obtaining a model with good generalization performance in target domain. We then propose variational bottleneck domain adaptation (VBDA), a new domain adaptation method which improves feature transferability by explicitly enforcing the feature extractor to ignore the task-irrelevant factors and focus on the information that is essential to the task of interest for both source and target domains. Extensive experimental results demonstrate that VBDA significantly outperforms state-of-the-art methods across three domain adaptation benchmark datasets.

Model-based reinforcement learning approaches leverage a forward dynamics model to support planning and decision making, which, however, may fail catastrophically if the model is inaccurate. Although there are several existing methods dedicated to combating the model error, the potential of the single forward model is still limited. In this paper, we propose to additionally construct a backward dynamics model to reduce the reliance on accuracy in forward model predictions. We develop a novel method, called Bidirectional Model-based Policy Optimization (BMPO) to utilize both the forward model and backward model to generate short branched rollouts for policy optimization. Furthermore, we theoretically derive a tighter bound of return discrepancy, which shows the superiority of BMPO against the one using merely the forward model. Extensive experiments demonstrate that BMPO outperforms state-of-the-art model-based methods in terms of sample efficiency and asymptotic performance.

Centralized training with decentralized execution has become an important paradigm in multi-agent learning. Though practical, current methods rely on restrictive assumptions to decompose the centralized value function across agents for execution. In this paper, we eliminate this restriction by proposing multi-agent determinantal Q-learning. Our method is established on Q-DPP, a novel extension of determinantal point process (DPP) to multi-agent setting. Q-DPP promotes agents to acquire diverse behavioral models; this allows a natural factorization of the joint Q-functions with no need for \emph{a priori} structural constraints on the value function or special network architectures. We demonstrate that Q-DPP generalizes major solutions including VDN, QMIX, and QTRAN on decentralizable cooperative tasks. To efficiently draw samples from Q-DPP, we develop a linear-time sampler with theoretical approximation guarantee. Our sampler also benefits exploration by coordinating agents to cover orthogonal directions in the state space during training. We evaluate our algorithm on multiple cooperative benchmarks; its effectiveness has been demonstrated when compared with the state-of-the-art.

The linear contextual bandits is a sequential decision-making problem where an agent decides among sequential actions given their corresponding contexts. Since large-scale data sets become more and more common, we study the linear contextual bandits in high-dimensional situations. Recent works focus on employing matrix sketching methods to accelerating contextual bandits. However, the matrix approximation error will bring additional terms to the regret bound. In this paper we first propose a novel matrix sketching method which is called Spectral Compensation Frequent Directions (SCFD). Then we propose an efficient approach for contextual bandits by adopting SCFD to approximate the covariance matrices. By maintaining and manipulating sketched matrices, our method only needs O(md) space and O(md) updating time in each round, where d is the dimensionality of the data and m is the sketching size. Theoretical analysis reveals that our method has better regret bounds than previous methods in high-dimensional cases. Experimental results demonstrate the effectiveness of our algorithm and verify our theoretical guarantees.

Neural architecture search (NAS) has shown encouraging results in automating the architecture design. Recently, DARTS relaxes the search process with a differentiable formulation that leverages weight-sharing and SGD for cost reduction of NAS. In DARTS, all candidate operations are trained simultaneously during the network weight training step. Our empirical results show that this training procedure leads to the co-adaption problem and Matthew Effect: operations with fewer parameters would be trained maturely earlier. This causes two problems: firstly, the operations with more parameters may never have the chance to express the desired function since those with less have already done the job; secondly, the system will punish those underperforming operations by lowering their architecture parameter and backward smaller loss gradients, this causes the Matthew Effect. In this paper, we systematically study these problems and propose a novel grouped operation dropout algorithm named DropNAS to fix the problems with DARTS. Extensive experiments demonstrate that DropNAS solves the above issues and achieves promising performance. Specifically, DropNAS achieves 2.26% test error on CIFAR-10, 16.39% on CIFAR-100 and 23.4% on ImageNet (with the same training hyperparameters as DARTS for a fair comparison). It is also observed that DropNAS is robust across variants of the DARTS search space. Code is available at https://github.com/wiljohnhong/DropNAS.

Aggregating crowd wisdom infers true labels for objects, from multiple noisy labels provided by various sources. Besides labels from sources, side information such as object features is also introduced to achieve higher inference accuracy. Usually, the learning-from-crowds framework is adopted. However, the framework considers each object in isolation and does not make full use of object features to overcome label noise. In this paper, we propose a clustering-based label-aware autoencoder (CLA) to alleviate label noise. CLA utilizes clusters to gather objects with similar features and exploits clustering to infer true labels, by constructing a novel deep generative process to simultaneously generate object features and source labels from clusters. For model inference, CLA extends the framework of variational autoencoders and utilizes maximizing a posteriori (MAP) estimation, which prevents the model from overfitting and trivial solutions. Experiments on real-world tasks demonstrate the significant improvement of CLA compared with the state-of-the-art aggregation algorithms.

There is an influx of heterogeneous information network (HIN) based recommender systems in recent years since HIN is capable of characterizing complex graphs and contains rich semantics. Although the existing approaches have achieved performance improvement, while practical, they still face the following problems. On one hand, most existing HIN-based methods rely on explicit path reachability to leverage path-based semantic relatedness between users and items, e.g., metapath-based similarities. These methods are hard to use and integrate since path connections are sparse or noisy, and are often of different lengths. On the other hand, other graph-based methods aim to learn effective heterogeneous network representations by compressing node together with its neighborhood information into single embedding before prediction. This weakly coupled manner in modeling overlooks the rich interactions among nodes, which introduces an early summarization issue. In this paper, we propose an end-to-end Neighborhood-based Interaction Model for Recommendation (NIRec) to address above problems. Specifically, we first analyze the significance of learning interactions in HINs and then propose a novel formulation to capture the interactive patterns between each pair of nodes through their metapath-guided neighborhoods. Then, to explore complex interactions between metapaths and deal with the learning complexity on large-scale networks, we formulate interaction in a convolutional way and learn efficiently with fast Fourier transform. The extensive experiments on four different types of heterogeneous graphs demonstrate the performance gains of NIRec comparing with state-of-the-arts. To the best of our knowledge, this is the first work providing an efficient neighborhood-based interaction model in the HIN-based recommendations.

Learning feature interactions is crucial for click-through rate (CTR) prediction in recommender systems. In most existing deep learning models, feature interactions are either manually designed or simply enumerated. However, enumerating all feature interactions brings large memory and computation cost. Even worse, useless interactions may introduce noise and complicate the training process. In this work, we propose a two-stage algorithm called Automatic Feature Interaction Selection (AutoFIS). AutoFIS can automatically identify important feature interactions for factorization models with computational cost just equivalent to training the target model to convergence. In the search stage, instead of searching over a discrete set of candidate feature interactions, we relax the choices to be continuous by introducing the architecture parameters. By implementing a regularized optimizer over the architecture parameters, the model can automatically identify and remove the redundant feature interactions during the training process of the model. In the re-train stage, we keep the architecture parameters serving as an attention unit to further boost the performance. Offline experiments on three large-scale datasets (two public benchmarks, one private) demonstrate that AutoFIS can significantly improve various FM based models. AutoFIS has been deployed onto the training platform of Huawei App Store recommendation service, where a 10-day online A/B test demonstrated that AutoFIS improved the DeepFM model by 20.3% and 20.1% in terms of CTR and CVR respectively.

Model-based reinforcement learning methods learn a dynamics model with real data sampled from the environment and leverage it to generate simulated data to derive an agent. However, due to the potential distribution mismatch between simulated data and real data, this could lead to degraded performance. Despite much effort being devoted to reducing this distribution mismatch, existing methods fail to solve it explicitly. In this paper, we investigate how to bridge the gap between real and simulated data due to inaccurate model estimation for better policy optimization. To begin with, we first derive a lower bound of the expected return, which naturally inspires a bound maximization algorithm by aligning the simulated and real data distributions. To this end, we propose a novel model-based reinforcement learning framework AMPO, which introduces unsupervised model adaptation to minimize the integral probability metric (IPM) between feature distributions from real and simulated data. Instantiating our framework with Wasserstein-1 distance gives a practical model-based approach. Empirically, our approach achieves state-of-the-art performance in terms of sample efficiency on a range of continuous control benchmark tasks.

The Frank-Wolfe algorithm is a classic method for constrained optimization problems. It has recently been popular in many machine learning applications because its projection-free property leads to more efficient iterations. In this paper, we study projection-free algorithms for convex-strongly-concave saddle point problems with complicated constraints. Our method combines Conditional Gradient Sliding with Mirror-Prox and show that it only requires $\tilde{\cO}(1/\sqrt{\epsilon})$ gradient evaluations and $\tilde{\cO}(1/\epsilon^2)$ linear optimizations in the batch setting. We also extend our method to the stochastic setting and propose first stochastic projection-free algorithms for saddle point problems. Experimental results demonstrate the effectiveness of our algorithms and verify our theoretical guarantees.

Information retrieval (IR) techniques, such as search, recommendation and online advertising, satisfying users' information needs by suggesting users personalized objects (information or services) at the appropriate time and place, play a crucial role in mitigating the information overload problem. Since the widely use of mobile applications, more and more information retrieval services have provided interactive functionality and products. Thus, learning from interaction becomes a crucial machine learning paradigm for interactive IR, which is based on reinforcement learning. With recent great advances in deep reinforcement learning (DRL), there have been increasing interests in developing DRL based information retrieval techniques, which could continuously update the information retrieval strategies according to users' real-time feedback, and optimize the expected cumulative long-term satisfaction from users. Our workshop aims to provide a venue, which can bring together academia researchers and industry practitioners (i) to discuss the principles, limitations and applications of DRL for information retrieval, and (ii) to foster research on innovative algorithms, novel techniques, and new applications of DRL to information retrieval.

Position bias is a critical problem in information retrieval when dealing with implicit yet biased user feedback data. Unbiased ranking methods typically rely on causality models and debias the user feedback through inverse propensity weighting. While practical, these methods still suffer from two major problems. First, when infer a user click, the impact of the contextual information, such as documents that have been examined, is often ignored. Second, only the position bias is considered but other issues resulted from user browsing behaviors are overlooked. In this paper, we propose an end-to-end Deep Recurrent Survival Ranking (DRSR), a unified framework to jointly model user's various behaviors, to (i) consider the rich contextual information in the ranking list; and (ii) address the hidden issues underlying user behaviors, i.e., to mine observe pattern in queries without any click (non-click queries), and to model tracking logs which cannot truly reflect the user browsing intents (untrusted observation). Specifically, we adopt a recurrent neural network to model the contextual information and estimates the conditional likelihood of user feedback at each position. We then incorporate survival analysis techniques with the probability chain rule to mathematically recover the unbiased joint probability of one user's various behaviors. DRSR can be easily incorporated with both point-wise and pair-wise learning objectives. The extensive experiments over two large-scale industrial datasets demonstrate the significant performance gains of our model comparing with the state-of-the-arts.

Click-through rate (CTR) prediction plays a key role in modern online personalization services. In practice, it is necessary to capture user's drifting interests by modeling sequential user behaviors to build an accurate CTR prediction model. However, as the users accumulate more and more behavioral data on the platforms, it becomes non-trivial for the sequential models to make use of the whole behavior history of each user. First, directly feeding the long behavior sequence will make online inference time and system load infeasible. Second, there is much noise in such long histories to fail the sequential model learning. The current industrial solutions mainly truncate the sequences and just feed recent behaviors to the prediction model, which leads to a problem that sequential patterns such as periodicity or long-term dependency are not embedded in the recent several behaviors but in far back history. To tackle these issues, in this paper we consider it from the data perspective instead of just designing more sophisticated yet complicated models and propose User Behavior Retrieval for CTR prediction (UBR4CTR) framework. In UBR4CTR, the most relevant and appropriate user behaviors will be firstly retrieved from the entire user history sequence using a learnable search method. These retrieved behaviors are then fed into a deep model to make the final prediction instead of simply using the most recent ones. It is highly feasible to deploy UBR4CTR into industrial model pipeline with low cost. Experiments on three real-world large-scale datasets demonstrate the superiority and efficacy of our proposed framework and models.

In this paper, we present the design of QuAChIE, a Question Answering based Chinese Information Extraction system. QuAChIE mainly depends on a well-trained question answering model to extract high-quality triples. The group of head entity and relation are regarded as a question given the input text as the context. For the training and evaluation of each model in the system, we build a large-scale information extraction dataset using Wikidata and Wikipedia pages by distant supervision. The advanced models implemented on top of the pre-trained language model and the enormous distant supervision data enable QuAChIE to extract relation triples from documents with cross-sentence correlations. The experimental results on the test set and the case study based on the interactive demonstration show its satisfactory Information Extraction quality on Chinese document-level texts.

Spaced repetition technique aims at improving long-term memory retention for human students by exploiting repeated, spaced reviews of learning contents. The study of spaced repetition focuses on designing an optimal policy to schedule the learning contents. To the best of our knowledge, none of the existing methods based on reinforcement learning take into account the varying time intervals between two adjacent learning events of the student, which, however, are essential to determine real-world schedule. In this paper, we aim to learn a scheduling policy that fully exploits the varying time interval information with high sample efficiency. We propose the Time-Aware scheduler with Dyna-Style planning (TADS) approach: a sample-efficient reinforcement learning framework for realistic spaced repetition. TADS learns a Time-LSTM policy to select an optimal content according to the student's whole learning history and the time interval since the last learning event. Besides, Dyna-style planning is integrated into TADS to further improve the sample efficiency. We evaluate our approach on three environments built from synthetic data and real-world data based on well-recognized cognitive models. Empirical results demonstrate that TADS achieves superior performance against state-of-the-art algorithms.

Interactive recommender system (IRS) has drawn huge attention because of its flexible recommendation strategy and the consideration of optimal long-term user experiences. To deal with the dynamic user preference and optimize accumulative utilities, researchers have introduced reinforcement learning (RL) into IRS. However, RL methods share a common issue of sample efficiency, i.e., huge amount of interaction data is required to train an effective recommendation policy, which is caused by the sparse user responses and the large action space consisting of a large number of candidate items. Moreover, it is infeasible to collect much data with explorative policies in online environments, which will probably harm user experience. In this work, we investigate the potential of leveraging knowledge graph (KG) in dealing with these issues of RL methods for IRS, which provides rich side information for recommendation decision making. Instead of learning RL policies from scratch, we make use of the prior knowledge of the item correlation learned from KG to (i) guide the candidate selection for better candidate item retrieval, (ii) enrich the representation of items and user states, and (iii) propagate user preferences among the correlated items over KG to deal with the sparsity of user feedback. Comprehensive experiments have been conducted on two real-world datasets, which demonstrate the superiority of our approach with significant improvements against state-of-the-arts.

Sequential recommendation task aims to predict user preference over items in the future given user historical behaviors. The order of user behaviors implies that there are resourceful sequential patterns embedded in the behavior history which reveal the underlying dynamics of user interests. Various sequential recommendation methods are proposed to model the dynamic user behaviors. However, most of the models only consider the user's own behaviors and dynamics, while ignoring the collaborative relations among users and items, i.e., similar tastes of users or analogous properties of items. Without modeling collaborative relations, those methods suffer from the lack of recommendation diversity and thus may have worse performance. Worse still, most existing methods only consider the user-side sequence and ignore the temporal dynamics on the item side. To tackle the problems of the current sequential recommendation models, we propose Sequential Collaborative Recommender (SCoRe) which effectively mines high-order collaborative information using cross-neighbor relation modeling and, additionally utilizes both user-side and item-side historical sequences to better capture user and item dynamics. Experiments on three real-world yet large-scale datasets demonstrate the superiority of the proposed model over strong baselines.

Reinforcement learning (RL) has recently been introduced to interactive recommender systems (IRS) because of its nature of learning from dynamic interactions and planning for long-run performance. As IRS is always with thousands of items to recommend (i.e., thousands of actions), most existing RL-based methods, however, fail to handle such a large discrete action space problem and thus become inefficient. The existing work that tries to deal with the large discrete action space problem by utilizing the deep deterministic policy gradient framework suffers from the inconsistency between the continuous action representation (the output of the actor network) and the real discrete action. To avoid such inconsistency and achieve high efficiency and recommendation effectiveness, in this paper, we propose a Tree-structured Policy Gradient Recommendation (TPGR) framework, where a balanced hierarchical clustering tree is built over the items and picking an item is formulated as seeking a path from the root to a certain leaf of the tree. Extensive experiments on carefully-designed environments based on two real-world datasets demonstrate that our model provides superior recommendation performance and significant efficiency improvement over state-of-the-art methods.

We present Academic Reader, a system which can read academic literatures and answer the relevant questions for researchers. Academic Reader leverages machine reading comprehension technique, which has been successfully applied in many fields but has not been involved in academic literature reading. An interactive platform is established to demonstrate the functions of Academic Reader. Pieces of academic literature and relevant questions are input to our system, which then outputs answers. The system can also gather users’ revised answers and perform active learning to continuously improve its performance. A case study is provided presenting the performance of our system on all papers accepted in KDD 2018, which demonstrates how our system facilitates massive academic literature reading.

Survival analysis is a hotspot in statistical research for modeling time-to-event information with data censorship handling, which has been widely used in many applications such as clinical research, information system and other fields with survivorship bias. Many works have been proposed for survival analysis ranging from traditional statistic methods to machine learning models. However, the existing methodologies either utilize counting-based statistics on the segmented data, or have a pre-assumption on the event probability distribution w.r.t. time. Moreover, few works consider sequential patterns within the feature space. In this paper, we propose a Deep Recurrent Survival Analysis model which combines deep learning for conditional probability prediction at finegrained level of the data, and survival analysis for tackling the censorship. By capturing the time dependency through modeling the conditional probability of the event for each sample, our method predicts the likelihood of the true event occurrence and estimates the survival rate over time, i.e., the probability of the non-occurrence of the event, for the censored data. Meanwhile, without assuming any specific form of the event probability distribution, our model shows great advantages over the previous works on fitting various sophisticated data distributions. In the experiments on the three realworld tasks from different fields, our model significantly outperforms the state-of-the-art solutions under various metrics.

Text-based question answering (TBQA) has been studied extensively in recent years.Most existing approaches focus on finding the answer to a question within a single paragraph.However, many difficult questions require multiple supporting evidence from scattered text across two or more documents.In this paper, we propose the Dynamically Fused Graph Network (DFGN), a novel method to answer those questions requiring multiple scattered evidence and reasoning over them.Inspired by human's step-by-step reasoning behavior, DFGN includes a dynamic fusion layer that starts from the entities mentioned in the given query, explores along the entity graph dynamically built from the text, and gradually finds relevant supporting entities from the given documents.We evaluate DFGN on HotpotQA, a public TBQA dataset requiring multi-hop reasoning.DFGN achieves competitive results on the public board.Furthermore, our analysis shows DFGN could produce interpretable reasoning chains.

Data-enabled smart transportation has attracted a surge of interest from machine learning and data mining researchers nowadays due to the bloom of online ride-hailing industry and rapid development of autonomous driving. Large-scale high quality route data and trading data (spatiotemporal data) have been generated every day, which makes AI an urgent need and preferred solution for the decision making in intelligent transportation systems. While a large of amount of work have been dedicated to traditional transportation problems, they are far from satisfactory for the rising need. We propose a half-day workshop at CIKM 2019 for the professionals, researchers, and practitioners who are interested in mining and understanding big and heterogeneous data generated in transportation, and AI applications to improve the transportation system. We plan to have several invited talks from both academia and industry. This workshop would be organized by Shanghai Jiao Tong University, Didi Chuxing and Pennsylvania State University.

Most e-commerce product feeds provide blended results of advertised products and recommended products to consumers. The underlying advertising and recommendation platforms share similar if not exactly the same set of candidate products. Consumers' behaviors on the advertised results constitute part of the recommendation model's training data and therefore can influence the recommended results. We refer to this process as Leverage. Considering this mechanism, we propose a novel perspective that advertisers can strategically bid through the advertising platform to optimize their recommended organic traffic. By analyzing the real-world data, we first explain the principles of Leverage mechanism, i.e., the dynamic models of Leverage. Then we introduce a novel Leverage optimization problem and formulate it with a Markov Decision Process. To deal with the sample complexity challenge in model-free reinforcement learning, we propose a novel Hybrid Training Leverage Bidding (HTLB) algorithm which combines the real-world samples and the emulator-generated samples to boost the learning speed and stability. Our offline experiments as well as the results from the online deployment demonstrate the superior performance of our approach.

How to optimally dispatch orders to vehicles and how to trade off between immediate and future returns are fundamental questions for a typical ride-hailing platform. We model ride-hailing as a large-scale parallel ranking problem and study the joint decision-making task of order dispatching and fleet management in online ride-hailing platforms. This task brings unique challenges in the following four aspects. First, to facilitate a huge number of vehicles to act and learn efficiently and robustly, we treat each region cell as an agent and build a multi-agent reinforcement learning framework. Second, to coordinate the agents from different regions to achieve long-term benefits, we leverage the geographical hierarchy of the region grids to perform hierarchical reinforcement learning. Third, to deal with the heterogeneous and variant action space for joint order dispatching and fleet management, we design the action as the ranking weight vector to rank and select the specific order or the fleet management destination in a unified formulation. Fourth, to achieve the multi-scale ride-hailing platform, we conduct the decision-making process in a hierarchical way where a multi-head attention mechanism is utilized to incorporate the impacts of neighbor agents and capture the key agent in each scale. The whole novel framework is named as CoRide. Extensive experiments based on multiple cities real-world data as well as analytic synthetic data demonstrate that CoRide provides superior performance in terms of platform revenue and user experience in the task of city-wide hybrid order dispatching and fleet management over strong baselines.

In E-commerce advertising, where product recommendations and product ads are presented to users simultaneously, the traditional setting is to display ads at fixed positions. However, under such a setting, the advertising system loses the flexibility to control the number and positions of ads, resulting in sub-optimal platform revenue and user experience. Consequently, major e-commerce platforms (e.g., Taobao.com) have begun to consider more flexible ways to display ads. In this paper, we investigate the problem of advertising with adaptive exposure: can we dynamically determine the number and positions of ads for each user visit under certain business constraints so that the platform revenue can be increased? More specifically, we consider two types of constraints: request-level constraint ensures user experience for each user visit, and platform-level constraint controls the overall platform monetization rate. We model this problem as a Constrained Markov Decision Process with per-state constraint (psCMDP) and propose a constrained two-level reinforcement learning approach to decompose the original problem into two relatively independent sub-problems. To accelerate policy learning, we also devise a constrained hindsight experience replay mechanism. Experimental evaluations on industry-scale real-world datasets demonstrate the merits of our approach in both obtaining higher revenue under the constraints and the effectiveness of the constrained hindsight experience replay mechanism.

Cooperation among the traffic signals enables vehicles to move through intersections more quickly. Conventional transportation approaches implement cooperation by pre-calculating the offsets between two intersections. Such pre-calculated offsets are not suitable for dynamic traffic environments. To enable cooperation of traffic signals, in this paper, we propose a model, CoLight, which uses graph attentional networks to facilitate communication. Specifically, for a target intersection in a network, CoLight can not only incorporate the temporal and spatial influences of neighboring intersections to the target intersection, but also build up index-free modeling of neighboring intersections. To the best of our knowledge, we are the first to use graph attentional networks in the setting of reinforcement learning for traffic signal control and to conduct experiments on the large-scale road network with hundreds of traffic signals. In experiments, we demonstrate that by learning the communication, the proposed model can achieve superior performance against the state-of-the-art methods.

Improving the efficiency of dispatching orders to vehicles is a research hotspot in online ride-hailing systems. Most of the existing solutions for order-dispatching are centralized controlling, which require to consider all possible matches between available orders and vehicles. For large-scale ride-sharing platforms, there are thousands of vehicles and orders to be matched at every second which is of very high computational cost. In this paper, we propose a decentralized execution order-dispatching method based on multi-agent reinforcement learning to address the large-scale order-dispatching problem. Different from the previous cooperative multi-agent reinforcement learning algorithms, in our method, all agents work independently with the guidance from an evaluation of the joint policy since there is no need for communication or explicit cooperation between agents. Furthermore, we use KL-divergence optimization at each time step to speed up the learning process and to balance the vehicles (supply) and orders (demand). Experiments on both the explanatory environment and real-world simulator show that the proposed method outperforms the baselines in terms of accumulated driver income (ADI) and Order Response Rate (ORR) in various traffic environments. Besides, with the support of the online platform of Didi Chuxing, we designed a hybrid system to deploy our model.

Exploration is crucial for training the optimal reinforcement learning (RL) policy, where the key is to discriminate whether a state visiting is novel. Most previous work focuses on designing heuristic rules or distance metrics to check whether a state is novel without considering such a discrimination process that can be learned. In this paper, we propose a novel method called generative adversarial exploration (GAEX) to encourage exploration in RL via introducing an intrinsic reward output from a generative adversarial network, where the generator provides fake samples of states that help discriminator identify those less frequently visited states. Thus the agent is encouraged to visit those states which the discriminator is less confident to judge as visited. GAEX is easy to implement and of high training efficiency. In our experiments, we apply GAEX into DQN and the DQN-GAEX algorithm achieves convincing performance on challenging exploration problems, including the game Venture, Montezuma's Revenge and Super Mario Bros, without further fine-tuning on complicate learning algorithms. To our knowledge, this is the first work to employ GAN in RL exploration problems.

Deep Q-learning has achieved significant success in single-agent decision making tasks. However, it is challenging to extend Q-learning to large-scale multi-agent scenarios, due to the explosion of action space resulting from the complex dynamics between the environment and the agents. In this paper, we propose to make the computation of multi-agent Q-learning tractable by treating the Q-function (w.r.t. state and joint-action) as a high-order high-dimensional tensor and then approximate it with factorized pairwise interactions. Furthermore, we utilize a composite deep neural network architecture for computing the factorized Q-function, share the model parameters among all the agents within the same group, and estimate the agents' optimal joint actions through a coordinate descent type algorithm. All these simplifications greatly reduce the model complexity and accelerate the learning process. Extensive experiments on two different multi-agent problems demonstrate the performance gain of our proposed approach in comparison with strong baselines, particularly when there are a large number of agents.

Lihua Qian, Lin Qiu, Weinan Zhang, Xin Jiang, Yong Yu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

In this paper, we study the generative models of sequential discrete data. To tackle the exposure bias problem inherent in maximum likelihood estimation (MLE), generative adversarial networks (GANs) are introduced to penalize the unrealistic generated samples. To exploit the supervision signal from the discriminator, most previous models leverage REINFORCE to address the non-differentiable problem of sequential discrete data. However, because of the unstable property of the training signal during the dynamic process of adversarial training, the effectiveness of REINFORCE, in this case, is hardly guaranteed. To deal with such a problem, we propose a novel approach called Cooperative Training (CoT) to improve the training of sequence generative models. CoT transforms the min-max game of GANs into a joint maximization framework and manages to explicitly estimate and optimize Jensen-Shannon divergence. Moreover, CoT works without the necessity of pre-training via MLE, which is crucial to the success of previous methods. In the experiments, compared to existing state-of-the-art methods, CoT shows superior or at least competitive performance on sample quality, diversity, as well as training stability.

In this paper we show that generative adversarial networks (GANs) without restriction on the discriminative function space commonly suffer from the problem that the gradient produced by the discriminator is uninformative to guide the generator. By contrast, Wasserstein GAN (WGAN), where the discriminative function is restricted to 1-Lipschitz, does not suffer from such a gradient uninformativeness problem. We further show in the paper that the model with a compact dual form of Wasserstein distance, where the Lipschitz condition is relaxed, may also theoretically suffer from this issue. This implies the importance of Lipschitz condition and motivates us to study the general formulation of GANs with Lipschitz constraint, which leads to a new family of GANs that we call Lipschitz GANs (LGANs). We show that LGANs guarantee the existence and uniqueness of the optimal discriminative function as well as the existence of a unique Nash equilibrium. We prove that LGANs are generally capable of eliminating the gradient uninformativeness problem. According to our empirical analysis, LGANs are more stable and generate consistently higher quality samples compared with WGAN.

In this paper we propose a hybrid architecture of actor-critic algorithms for reinforcement learning in parameterized action space, which consists of multiple parallel sub-actor networks to decompose the structured action space into simpler action spaces along with a critic network to guide the training of all sub-actor networks. While this paper is mainly focused on parameterized action space, the proposed architecture, which we call hybrid actor-critic, can be extended for more general action spaces which has a hierarchical structure. We present an instance of the hybrid actor-critic architecture based on proximal policy optimization (PPO), which we refer to as hybrid proximal policy optimization (H-PPO). Our experiments test H-PPO on a collection of tasks with parameterized action space, where H-PPO demonstrates superior performance over previous methods of parameterized action reinforcement learning.

Trajectory data has been widely used in many urban applications. Sharing trajectory data with effective supervision is a vital task, as it contains private information of moving objects. However, malicious data users can modify trajectories in various ways to avoid data distribution tracking by the hashing-based data signatures, e.g., MD5. Moreover, the existing trajectory data protection scheme can only protect trajectories from either spatial or temporal modifications. Finally, so far there is no authoritative third party for trajectory data sharing process, as trajectory data is too sensitive. To this end, we propose a novel trajectory copyright protection scheme, which can protect trajectory data from comprehensive types of data modifications/attacks. Three main techniques are employed to effectively guarantee the robustness and comprehensiveness of the proposed data sharing scheme: 1) the identity information is embedded distributively across a set of sub-trajectories partitioned based on the spatio-temporal regions; 2) the centroid distance of the sub-trajectories is served as a stable trajectory attribute to embed the information; and 3) the blockchain technique is used as a trusted third party to log all data transaction history for data distribution tracking in a decentralized manner. Extensive experiments were conducted based on two real-world trajectory datasets to demonstrate the effectiveness of our proposed scheme.

The emergence of real-time auction in online advertising has drawn huge attention of modeling the market competition, i.e., bid landscape forecasting. The problem is formulated as to forecast the probability distribution of market price for each ad auction. With the consideration of the censorship issue which is caused by the second-price auction mechanism, many researchers have devoted their efforts on bid landscape forecasting by incorporating survival analysis from medical research field. However, most existing solutions mainly focus on either counting-based statistics of the segmented sample clusters, or learning a parameterized model based on some heuristic assumptions of distribution forms. Moreover, they neither consider the sequential patterns of the feature over the price space. In order to capture more sophisticated yet flexible patterns at fine-grained level of the data, we propose a Deep Landscape Forecasting (DLF) model which combines deep learning for probability distribution forecasting and survival analysis for censorship handling. Specifically, we utilize a recurrent neural network to flexibly model the conditional winning probability w.r.t. each bid price. Then we conduct the bid landscape forecasting through probability chain rule with strict mathematical derivations. And, in an end-to-end manner, we optimize the model by minimizing two negative likelihood losses with comprehensive motivations. Without any specific assumption for the distribution form of bid landscape, our model shows great advantages over previous works on fitting various sophisticated market price distributions. In the experiments over two large-scale real-world datasets, our model significantly outperforms the state-of-the-art solutions under various metrics.

User response prediction, which models the user preference w.r.t. the presented items, plays a key role in online services. With two-decade rapid development, nowadays the cumulated user behavior sequences on mature Internet service platforms have become extremely long since the user's first registration. Each user not only has intrinsic tastes, but also keeps changing her personal interests during lifetime. Hence, it is challenging to handle such lifelong sequential modeling for each individual user. Existing methodologies for sequential modeling are only capable of dealing with relatively recent user behaviors, which leaves huge space for modeling long-term especially lifelong sequential patterns to facilitate user modeling. Moreover, one user's behavior may be accounted for various previous behaviors within her whole online activity history, i.e., long-term dependency with multi-scale sequential patterns. In order to tackle these challenges, in this paper, we propose a Hierarchical Periodic Memory Network for lifelong sequential modeling with personalized memorization of sequential patterns for each user. The model also adopts a hierarchical and periodical updating mechanism to capture multi-scale sequential patterns of user interests while supporting the evolving user behavior logs. The experimental results over three large-scale real-world datasets have demonstrated the advantages of our proposed model with significant improvement in user response prediction performance against the state-of-the-arts.

Knowledge base is one of the main forms to represent information in a structured way. A knowledge base typically consists of Resource Description Frameworks (RDF) triples which describe the entities and their relations. Generating natural language description of the knowledge base is an important task in NLP, which has been formulated as a conditional language generation task and tackled using the sequence-to-sequence framework. Current works mostly train the language models by maximum likelihood estimation, which tends to generate lousy sentences. In this paper, we argue that such a problem of maximum likelihood estimation is intrinsic, which is generally irrevocable via changing network structures. Accordingly, we propose a novel Triple-to-Text (T2T) framework, which approximately optimizes the inverse Kullback-Leibler (KL) divergence between the distributions of the real and generated sentences. Due to the nature that inverse KL imposes large penalty on fake-looking samples, the proposed method can significantly reduce the probability of generating low-quality sentences. Our experiments on three real-world datasets demonstrate that T2T can generate higher-quality sentences and outperform baseline models in several evaluation metrics.

In graphs with rich texts, incorporating textual information with structural information would benefit constructing expressive graph embeddings. Among various graph embedding models, random walk (RW)-based is one of the most popular and successful groups. However, it is challenged by two issues when applied on graphs with rich texts: (i) sampling efficiency: deriving from the training objective of RW-based models (e.g., DeepWalk and node2vec), we show that RW-based models are likely to generate large amounts of redundant training samples due to three main drawbacks. (ii) text utilization: these models have difficulty in dealing with zero-shot scenarios where graph embedding models have to infer graph structures directly from texts. To solve these problems, we propose a novel framework, namely Text-driven Graph Embedding with Pairs Sampling (TGE-PS). TGE-PS uses Pairs Sampling (PS) to improve the sampling strategy of RW, being able to reduce ~ 99% training samples while preserving competitive performance. TGE-PS uses Text-driven Graph Embedding (TGE), an inductive graph embedding approach, to generate node embeddings from texts. Since each node contains rich texts, TGE is able to generate high-quality embeddings and provide reasonable predictions on existence of links to unseen nodes. We evaluate TGE-PS on several real-world datasets, and experiment results demonstrate that TGE-PS produces state-of-the-art results on both traditional and zero-shot link prediction tasks.

Community detection refers to the task of discovering groups of vertices sharing similar properties or functions so as to understand the network data. With the recent development of deep learning, graph representation learning techniques are also utilized for community detection. However, the communities can only be inferred by applying clustering algorithms based on learned vertex embeddings. These general cluster algorithms like K-means and Gaussian Mixture Model cannot output much overlapped communities, which have been proved to be very common in many real-world networks. In this paper, we propose CommunityGAN, a novel community detection framework that jointly solves overlapping community detection and graph representation learning. First, unlike the embedding of conventional graph representation learning algorithms where the vector entry values have no specific meanings, the embedding of CommunityGAN indicates the membership strength of vertices to communities. Second, a specifically designed Generative Adversarial Net (GAN) is adopted to optimize such embedding. Through the minimax competition between the motif-level generator and discriminator, both of them can alternatively and iteratively boost their performance and finally output a better community structure. Extensive experiments on synthetic data and real-world tasks demonstrate that CommunityGAN achieves substantial community detection performance gains over the state-of-the-art methods.

Traffic signal control is an emerging application scenario for reinforcement learning. Besides being as an important problem that affects people's daily life in commuting, traffic signal control poses its unique challenges for reinforcement learning in terms of adapting to dynamic traffic environment and coordinating thousands of agents including vehicles and pedestrians. A key factor in the success of modern reinforcement learning relies on a good simulator to generate a large number of data samples for learning. The most commonly used open-source traffic simulator SUMO is, however, not scalable to large road network and large traffic flow, which hinders the study of reinforcement learning on traffic scenarios. This motivates us to create a new traffic simulator CityFlow with fundamentally optimized data structures and efficient algorithms. CityFlow can support flexible definitions for road network and traffic flow based on synthetic and real-world data. It also provides user-friendly interface for reinforcement learning. Most importantly, CityFlow is more than twenty times faster than SUMO and is capable of supporting city-wide traffic simulation with an interactive render for monitoring. Besides traffic signal control, CityFlow could serve as the base for other transportation studies and can create new possibilities to test machine learning methods in the intelligent transportation domain.

User response prediction is a crucial component for personalized information retrieval and filtering scenarios, such as recommender system and web search. The data in user response prediction is mostly in a multi-field categorical format and transformed into sparse representations via one-hot encoding. Due to the sparsity problems in representation and optimization, most research focuses on feature engineering and shallow modeling. Recently, deep neural networks have attracted research attention on such a problem for their high capacity and end-to-end training scheme. In this article, we study user response prediction in the scenario of click prediction. We first analyze a coupled gradient issue in latent vector-based models and propose kernel product to learn field-aware feature interactions. Then, we discuss an insensitive gradient issue in DNN-based models and propose Product-based Neural Network, which adopts a feature extractor to explore feature interactions. Generalizing the kernel product to a net-in-net architecture, we further propose Product-network in Network (PIN), which can generalize previous models. Extensive experiments on four industrial datasets and one contest dataset demonstrate that our models consistently outperform eight baselines on both area under curve and log loss. Besides, PIN makes great click-through rate improvement (relatively 34.67%) in online A/B test.

Techniques for automatically designing deep neural network architectures such as reinforcement learning based approaches have recently shown promising results. However, their success is based on vast computational resources (e.g. hundreds of GPUs), making them difficult to be widely used. A noticeable limitation is that they still design and train each network from scratch during the exploration of the architecture space, which is highly inefficient. In this paper, we propose a new framework toward efficient architecture search by exploring the architecture space based on the current network and reusing its weights. We employ a reinforcement learning agent as the meta-controller, whose action is to grow the network depth or layer width with function-preserving transformations. As such, the previously validated networks can be reused for further exploration, thus saves a large amount of computational cost. We apply our method to explore the architecture space of the plain convolutional neural networks (no skip-connections, branching etc.) on image benchmark datasets (CIFAR-10, SVHN) with restricted computational resources (5 GPUs). Our method can design highly competitive networks that outperform existing networks using the same design scheme. On CIFAR-10, our model without skip-connections achieves 4.23% test error rate, exceeding a vast majority of modern architectures and approaching DenseNet. Furthermore, by applying our method to explore the DenseNet architecture space, we are able to achieve more accurate networks with fewer parameters.

Link prediction is a fundamental problem with a wide range of applications in various domains, which predicts the links that are not yet observed or the links that may appear in the future. Most existing works in this field only focus on modeling a single network, while real-world networks are actually aligned with each other. Network alignments contain valuable additional information for understanding the networks, and provide a new direction for addressing data insufficiency and alleviating cold start problem. However, there are rare works leveraging network alignments for better link prediction. Besides, neural network is widely employed in various domains while its capability of capturing high-level patterns and correlations for link prediction problem has not been adequately researched yet. Hence, in this paper we target atlink prediction over aligned networks using neural networks. The major challenge is the heterogeneousness of the considered networks, as the networks may have different characteristics, link purposes, etc. To overcome this, we propose a novel multi-neural-network framework MNN, where we have one individual neural network for each heterogeneous target or feature while the vertex representations are shared. We further discuss training methods for the multi-neural-network framework. Extensive experiments demonstrate that MNN outperforms the state-of-the-art methods and achieves 3% to 5% relative improvement of AUC score across different settings, particularly over 8% for cold start scenarios.

Automatically generating coherent and semantically meaningful text has many applications in machine translation, dialogue systems, image captioning, etc. Recently, by combining with policy gradient, Generative Adversarial Nets(GAN) that use a discriminative model to guide the training of the generative model as a reinforcement learning policy has shown promising results in text generation. However, the scalar guiding signal is only available after the entire text has been generated and lacks intermediate information about text structure during the generative process. As such, it limits its success when the length of the generated text samples is long (more than 20 words). In this paper, we propose a new framework, called LeakGAN, to address the problem for long text generation. We allow the discriminative net to leak its own high-level extracted features to the generative net to further help the guidance. The generator incorporates such informative signals into all generation steps through an additional MANAGER module, which takes the extracted features of current generated words and outputs a latent vector to guide the WORKER module for next-word generation.Our extensive experiments on synthetic data and various real-world tasks with Turing test demonstrate that LeakGAN is highly effective in long text generation and also improves the performance in short text generation scenarios. More importantly, without any supervision, LeakGAN would be able to implicitly learn sentence structures only through the interaction between MANAGER and WORKER.

In this paper, we show that the recent integration of statistical models with deep recurrent neural networks provides a new way of formulating volatility (the degree of variation of time series) models that have been widely used in time series analysis and prediction in finance. The model comprises a pair of complementary stochastic recurrent neural networks: the generative network models the joint distribution of the stochastic volatility process; the inference network approximates the conditional distribution of the latent variables given the observables. Our focus here is on the formulation of temporal dynamics of volatility over time under a stochastic recurrent neural network framework. Experiments on real-world stock price datasets demonstrate that the proposed model generates a better volatility estimation and prediction that outperforms mainstream methods, e.g., deterministic models such as GARCH and its variants, and stochastic models namely the MCMC-based stochvol as well as the Gaussian-process-based, on average negative log-likelihood.

Domain adaptation aims at generalizing a high-performance learner on a target domain via utilizing the knowledge distilled from a source domain which has a different but related data distribution. One solution to domain adaptation is to learn domain invariant feature representations while the learned representations should also be discriminative in prediction. To learn such representations, domain adaptation frameworks usually include a domain invariant representation learning approach to measure and reduce the domain discrepancy, as well as a discriminator for classification. Inspired by Wasserstein GAN, in this paper we propose a novel approach to learn domain invariant feature representations, namely Wasserstein Distance Guided Representation Learning (WDGRL). WDGRL utilizes a neural network, denoted by the domain critic, to estimate empirical Wasserstein distance between the source and target samples and optimizes the feature extractor network to minimize the estimated Wasserstein distance in an adversarial manner. The theoretical advantages of Wasserstein distance for domain adaptation lie in its gradient property and promising generalization bound. Empirical studies on common sentiment and image classification adaptation datasets demonstrate that our proposed WDGRL outperforms the state-of-the-art domain invariant representation learning approaches.

The goal of graph representation learning is to embed each vertex in a graph into a low-dimensional vector space. Existing graph representation learning methods can be classified into two categories: generative models that learn the underlying connectivity distribution in the graph, and discriminative models that predict the probability of edge existence between a pair of vertices. In this paper, we propose GraphGAN, an innovative graph representation learning framework unifying above two classes of methods, in which the generative model and discriminative model play a game-theoretical minimax game. Specifically, for a given vertex, the generative model tries to fit its underlying true connectivity distribution over all other vertices and produces "fake" samples to fool the discriminative model, while the discriminative model tries to detect whether the sampled vertex is from ground truth or generated by the generative model. With the competition between these two models, both of them can alternately and iteratively boost their performance. Moreover, when considering the implementation of generative model, we propose a novel graph softmax to overcome the limitations of traditional softmax function, which can be proven satisfying desirable properties of normalization, graph structure awareness, and computational efficiency. Through extensive experiments on real-world datasets, we demonstrate that GraphGAN achieves substantial gains in a variety of applications, including link prediction, node classification, and recommendation, over state-of-the-art baselines.

We introduce MAgent, a platform to support research and development of many-agent reinforcement learning. Unlike previous research platforms on single or multi-agent reinforcement learning, MAgent focuses on supporting the tasks and the applications that require hundreds to millions of agents. Within the interactions among a population of agents, it enables not only the study of learning algorithms for agents' optimal polices, but more importantly, the observation and understanding of individual agent's behaviors and social phenomena emerging from the AI society, including communication languages, leaderships, altruism. MAgent is highly scalable and can host up to one million agents on a single GPU server. MAgent also provides flexible configurations for AI researchers to design their customized environments and agents. In this demo, we present three environments designed on MAgent and show emerged collective intelligence by learning from scratch.

Real-time advertising allows advertisers to bid for each impression for a visiting user. To optimize specific goals such as maximizing revenue and return on investment (ROI) led by ad placements, advertisers not only need to estimate the relevance between the ads and user's interests, but most importantly require a strategic response with respect to other advertisers bidding in the market. In this paper, we formulate bidding optimization with multi-agent reinforcement learning. To deal with a large number of advertisers, we propose a clustering method and assign each cluster with a strategic bidding agent. A practical Distributed Coordinated Multi-Agent Bidding (DCMAB) has been proposed and implemented to balance the tradeoff between the competition and cooperation among advertisers. The empirical study on our industry-scaled real-world data has demonstrated the effectiveness of our methods. Our results show cluster-based bidding would largely outperform single-agent and bandit approaches, and the coordinated bidding achieves better overall objectives than purely self-interested bidding agents.

In online advertising, the Internet users may be exposed to a sequence of different ad campaigns, i.e., display ads, search, or referrals from multiple channels, before led up to any final sales conversion and transaction. For both campaigners and publishers, it is fundamentally critical to estimate the contribution from ad campaign touch-points during the customer journey (conversion funnel) and assign the right credit to the right ad exposure accordingly. However, the existing research on the multi-touch attribution problem lacks a principled way of utilizing the users' pre-conversion actions (i.e., clicks), and quite often fails to model the sequential patterns among the touch points from a user's behavior data. To make it worse, the current industry practice is merely employing a set of arbitrary rules as the attribution model, e.g., the popular last-touch model assigns 100% credit to the final touch-point regardless of actual attributions. In this paper, we propose a Dual-attention Recurrent Neural Network (DARNN) for the multi-touch attribution problem. It learns the attribution values through an attention mechanism directly from the conversion estimation objective. To achieve this, we utilize sequence-to-sequence prediction for user clicks, and combine both post-view and post-click attribution patterns together for the final conversion estimation. To quantitatively benchmark attribution models, we also propose a novel yet practical attribution evaluation scheme through the proxy of budget allocation (under the estimated attributions) over ad channels. The experimental results on two real datasets demonstrate the significant performance gains of our attribution model against the state of the art.

Most existing knowledge graphs (KGs) in academic domains suffer from problems of insufficient multi-relational information, name ambiguity and improper data format for large-scale machine processing. In this paper, we present AceKG, a new large-scale KG in academic domain. AceKG not only provides clean academic information, but also offers a large-scale benchmark dataset for researchers to conduct challenging data mining projects including link prediction, community detection and scholar classification. Specifically, AceKG describes 3.13 billion triples of academic facts based on a consistent ontology, including necessary properties of papers, authors, fields of study, venues and institutes, as well as the relations among them. To enrich the proposed knowledge graph, we also perform entity alignment with existing databases and rule-based inference. Based on AceKG, we conduct experiments of three typical academic data mining tasks and evaluate several state-of-the-art knowledge embedding and network representation learning approaches on the benchmark datasets built from AceKG. Finally, we discuss promising research directions that benefit from AceKG.

We introduce a new function-preserving transformation for efficient neural architecture search. This network transformation allows reusing previously trained networks and existing successful architectures that improves sample efficiency. We aim to address the limitation of current network transformation operations that can only perform layer-level architecture modifications, such as adding (pruning) filters or inserting (removing) a layer, which fails to change the topology of connection paths. Our proposed path-level transformation operations enable the meta-controller to modify the path topology of the given network while keeping the merits of reusing weights, and thus allow efficiently designing effective structures with complex path topologies like Inception models. We further propose a bidirectional tree-structured reinforcement learning meta-controller to explore a simple yet highly expressive tree-structured architecture space that can be viewed as a generalization of multi-branch architectures. We experimented on the image classification datasets with limited computational resources (about 200 GPU-hours), where we observed improved parameter efficiency and better test results (97.70% test accuracy on CIFAR-10 with 14.3M parameters and 74.6% top-1 accuracy on ImageNet in the mobile setting), demonstrating the effectiveness and transferability of our designed architectures.

Existing multi-agent reinforcement learning methods are limited typically to a small number of agents. When the agent number increases largely, the learning becomes intractable due to the curse of the dimensionality and the exponential growth of agent interactions. In this paper, we present Mean Field Reinforcement Learning where the interactions within the population of agents are approximated by those between a single agent and the average effect from the overall population or neighboring agents; the interplay between the two entities is mutually reinforced: the learning of the individual agent’s optimal policy depends on the dynamics of the population, while the dynamics of the population change according to the collective patterns of the individual policies. We develop practical mean field Q-learning and mean field Actor-Critic algorithms and analyze the convergence of the solution to Nash equilibrium. Experiments on Gaussian squeeze, Ising model, and battle games justify the learning effectiveness of our mean field approaches. In addition, we report the first result to solve the Ising model via model-free reinforcement learning methods.

In typical reinforcement learning (RL), the environment is assumed given and the goal of the learning is to identify an optimal policy for the agent taking actions through its interactions with the environment. In this paper, we extend this setting by considering the environment is not given, but controllable and learnable through its interaction with the agent at the same time. This extension is motivated by environment design scenarios in the real-world, including game design, shopping space design and traffic signal design. Theoretically, we find a dual Markov decision process (MDP) w.r.t. the environment to that w.r.t. the agent, and derive a policy gradient solution to optimizing the parametrized environment. Furthermore, discontinuous environments are addressed by a proposed general generative framework. Our experiments on a Maze game design task show the effectiveness of the proposed algorithms in generating diverse and challenging Mazes against various agent settings.

Zhenghui Wang, Yanru Qu, Liheng Chen, Jian Shen, Weinan Zhang, Shaodian Zhang, Yimei Gao, Gen Gu, Ken Chen, Yong Yu. Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers). 2018.

Generative adversarial nets (GANs) have been widely studied during the recent development of deep learning and unsupervised learning. With an adversarial training mechanism, GAN manages to train a generative model to fit the underlying unknown real data distribution under the guidance of the discriminative model estimating whether a data instance is real or generated. Such a framework is originally proposed for fitting continuous data distribution such as images, thus it is not straightforward to be directly applied to information retrieval scenarios where the data is mostly discrete, such as IDs, text and graphs. In this tutorial, we focus on discussing the GAN techniques and the variants on discrete data fitting in various information retrieval scenarios. (i) We introduce the fundamentals of GAN framework and its theoretic properties; (ii) we carefully study the promising solutions to extend GAN onto discrete data generation; (iii) we introduce IRGAN, the fundamental GAN framework of fitting single ID data distribution and the direct application on information retrieval; (iv) we further discuss the task of sequential discrete data generation tasks, e.g., text generation, and the corresponding GAN solutions; (v) we present the most recent work on graph/network data fitting with node embedding techniques by GANs. Meanwhile, we also introduce the relevant open-source platforms such as IRGAN and Texygen to help audience conduct research experiments on GANs in information retrieval. Finally, we conclude this tutorial with a comprehensive summarization and a prospect of further research directions for GANs in information retrieval.

We introduce Texygen, a benchmarking platform to support research on open-domain text generation models. Texygen has not only implemented a majority of text generation models, but also covered a set of metrics that evaluate the diversity, the quality and the consistency of the generated texts. The Texygen platform could help standardize the research on text generation and improve the reproductivity and reliability of future research work in text generation.

Malicious calls, i.e., telephony spams and scams, have been a long-standing challenging issue that causes billions of dollars of annual financial loss worldwide. This work presents the first machine learning-based solution without relying on any particular assumptions on the underlying telephony network infrastructures. The main challenge of this decade-long problem is that it is unclear how to construct effective features without the access to the telephony networks' infrastructures. We solve this problem by combining several innovations. We first develop a TouchPal user interface on top of a mobile App to allow users tagging malicious calls. This allows us to maintain a large-scale call log database. We then conduct a measurement study over three months of call logs, including 9 billion records. We design 29 features based on the results, so that machine learning algorithms can be used to predict malicious calls. We extensively evaluate different state-of-the-art machine learning approaches using the proposed features, and the results show that the best approach can reduce up to 90% unblocked malicious calls while maintaining a precision over 99.99% on the benign call traffic. The results also show the models are efficient to implement without incurring a significant latency overhead. We also conduct ablation analysis, which reveals that using 10 out of the 29 features can reach a performance comparable to using all features.

Although the word-popularity based negative sampler has shown superb performance in the skip-gram model, the theoretical motivation behind oversampling popular (non-observed) words as negative samples is still not well understood. In this paper, we start from an investigation of the gradient vanishing issue in the skip-gram model without a proper negative sampler. By performing an insightful analysis from the stochastic gradient descent (SGD) learning perspective, we demonstrate, both theoretically and intuitively, negative samples with larger inner product scores are more informative than those with lower scores for the SGD learner in terms of both convergence rate and accuracy. Understanding this, we propose an alternative sampling algorithm that dynamically selects informative negative samples during each SGD update. More importantly, the proposed sampler accounts for multi-dimensional self-embedded features during the sampling process, which essentially makes it more effective than the original popularity-based (one-dimensional) sampler. Empirical experiments further verify our observations and show that our fine-grained samplers gain significant improvement over the existing ones without increasing computational complexity.

Real-time bidding (RTB) based display advertising has become one of the key technological advances in computational advertising. RTB enables advertisers to buy individual ad impressions via an auction in real-time and facilitates the evaluation and the bidding of individual impressions across multiple advertisers. In RTB, the advertisers face three main challenges when optimizing their bidding strategies, namely (i) estimating the utility (e.g., conversions, clicks) of the ad impression, (ii) forecasting the market value (thus the cost) of the given ad impression, and (iii) deciding the optimal bid for the given auction based on the first two. Previous solutions assume the first two are solved before addressing the bid optimization problem. However, these challenges are strongly correlated and dealing with any individual problem independently may not be globally optimal. In this paper, we propose Bidding Machine, a comprehensive learning to bid framework, which consists of three optimizers dealing with each challenge above, and as a whole, jointly optimizes these three parts. We show that such a joint optimization would largely increase the campaign effectiveness and the profit. From the learning perspective, we show that the bidding machine can be updated smoothly with both offline periodical batch or online sequential training schemes. Our extensive offline empirical study and online A/B testing verify the high effectiveness of the proposed bidding machine.

As a new way of training generative models, Generative Adversarial Net (GAN) that uses a discriminative model to guide the training of the generative model has enjoyed considerable success in generating real-valued data. However, it has limitations when the goal is for generating sequences of discrete tokens. A major reason lies in that the discrete outputs from the generative model make it difficult to pass the gradient update from the discriminative model to the generative model. Also, the discriminative model can only assess a complete sequence, while for a partially generated sequence, it is non-trivial to balance its current score and the future one once the entire sequence has been generated. In this paper, we propose a sequence generation framework, called SeqGAN, to solve the problems. Modeling the data generator as a stochastic policy in reinforcement learning (RL), SeqGAN bypasses the generator differentiation problem by directly performing gradient policy update. The RL reward signal comes from the GAN discriminator judged on a complete sequence, and is passed back to the intermediate state-action steps using Monte Carlo search. Extensive experiments on synthetic data and real-world tasks demonstrate significant improvements over strong baselines.

Personalized recommendation has been proved effective as a content discovery tool for many online news publishers. As fresh news articles are frequently coming to the system while the old ones are fading away quickly, building a consistent and coherent feature representation over the ever-changing articles pool is fundamental to the performance of the recommendation. However, learning a good feature representation is challenging, especially for some small publishers that have normally fewer than 10,000 articles each year. In this paper, we consider to transfer knowledge from a larger text corpus. In our proposed solution, an effective article recommendation engine can be established with a small number of target publisher articles by transferring knowledge from a large corpus of text with a different distribution. Specifically, we leverage noise contrastive estimation techniques to learn the word conditional distribution given the context words, where the noise conditional distribution is pre-trained from the large corpus. Our solution has been deployed in a commercial recommendation service. The large-scale online A/B testing on two commercial publishers demonstrates up to 9.97% relative overall performance gain of our proposed model on the recommendation click-though rate metric over the non-transfer learning baselines.

Programmatic display advertising, which enables advertisers to make real-time decisions on individual ad display opportunities so as to achieve a precise audience marketing, has become a key technique for online advertising. However, the constrained budget setting still restricts unlimited ad impressions. As a result, a smart strategy for ad impression selection is necessary for the advertisers to maximize positive user responses such as clicks or conversions, under the constraints of both ad volume and campaign budget. In this paper, we borrow in the idea of top-N ranking and filtering techniques from information retrieval and propose an effective ad impression volume ranking method for each ad campaign, followed by a sequential selection strategy considering the remaining ad volume and budget, to smoothly deliver the volume filtering while maximizing campaign efficiency. The extensive experiments on two benchmarking datasets and a commercial ad platform demonstrate large performance superiority of our proposed solution over traditional methods, especially under tight budgets.

Aggregating crowd wisdoms takes multiple labels from various sources and infers true labels for objects. Recent research work makes progress by learning source credibility from data and roughly form three kinds of modeling frameworks: weighted majority voting, trust propagation, and generative models. In this paper, we propose a novel framework named Label-Aware Autoencoders (LAA) to aggregate crowd wisdoms. LAA integrates a classifier and a reconstructor into a unified model to infer labels in an unsupervised manner. Analogizing classical autoencoders, we can regard the classifier as an encoder, the reconstructor as a decoder, and inferred labels as latent features. To the best of our knowledge, it is the first trial to combine label aggregation with autoencoders. We adopt networks to implement the classifier and the reconstructor which have the potential to automatically learn underlying patterns of source credibility. To further improve inference accuracy, we introduce object ambiguity and latent aspects into LAA. Experiments on three real-world datasets show that proposed models achieve impressive inference accuracy improvement over state-of-the-art models.

Feature-based matrix factorization techniques such as Factorization Machines (FM) have been proven to achieve impressive accuracy for the rating prediction task. However, most common recommendation scenarios are formulated as a top-N item ranking problem with implicit feedback (e.g., clicks, purchases)rather than explicit ratings. To address this problem, with both implicit feedback and feature information, we propose a feature-based collaborative boosting recommender called BoostFM, which integrates boosting into factorization models during the process of item ranking. Specifically, BoostFM is an adaptive boosting framework that linearly combines multiple homogeneous component recommenders, which are repeatedly constructed on the basis of the individual FM model by a re-weighting scheme. Two ways are proposed to efficiently train the component recommenders from the perspectives of both pairwise and listwise Learning-to-Rank (L2R). The properties of our proposed method are empirically studied on three real-world datasets. The experimental results show that BoostFM outperforms a number of state-of-the-art approaches for top-N recommendation.

As aggregators, online news portals face great challenges in continuously selecting a pool of candidate articles to be shown to their users. Typically, those candidate articles are recommended manually by platform editors from a much larger pool of articles aggregated from multiple sources. Such a hand-pick process is labor intensive and time-consuming. In this paper, we study the editor article selection behavior and propose a learning by demonstration system to automatically select a subset of articles from the large pool. Our data analysis shows that (i) editors' selection criteria are non-explicit, which are less based only on the keywords or topics, but more depend on the quality and attractiveness of the writing from the candidate article, which is hard to capture based on traditional bag-of-words article representation. And (ii) editors' article selection behaviors are dynamic: articles with different data distribution come into the pool everyday and the editors' preference varies, which are driven by some underlying periodic or occasional patterns. To address such problems, we propose a meta-attention model across multiple deep neural nets to (i) automatically catch the editors' underlying selection criteria via the automatic representation learning of each article and its interaction with the meta data and (ii) adaptively capture the change of such criteria via a hybrid attention model. The attention model strategically incorporates multiple prediction models, which are trained in previous days. The system has been deployed in a commercial article feed platform. A 9-day A/B testing has demonstrated the consistent superiority of our proposed model over several strong baselines.

This paper provides a unified account of two schools of thinking in information retrieval modelling: the generative retrieval focusing on predicting relevant documents given a query, and the discriminative retrieval focusing on predicting relevancy given a query-document pair. We propose a game theoretical minimax game to iteratively optimise both models. On one hand, the discriminative model, aiming to mine signals from labelled and unlabelled data, provides guidance to train the generative model towards fitting the underlying relevance distribution over documents given the query. On the other hand, the generative model, acting as an attacker to the current discriminative model, generates difficult examples for the discriminative model in an adversarial way by minimising its discrimination objective. With the competition between these two models, we show that the unified framework takes advantage of both schools of thinking: (i) the generative model learns to fit the relevance distribution over documents via the signals from the discriminative model, and (ii) the discriminative model is able to exploit the unlabelled data selected by the generative model to achieve a better estimation for document ranking. Our experimental results have demonstrated significant performance gains as much as 23.96% on [email protected] and 15.50% on MAP over strong baselines in a variety of applications including web search, item recommendation, and question answering.

The majority of online display ads are served through real-time bidding (RTB) --- each ad display impression is auctioned off in real-time when it is just being generated from a user visit. To place an ad automatically and optimally, it is critical for advertisers to devise a learning algorithm to cleverly bid an ad impression in real-time. Most previous works consider the bid decision as a static optimization problem of either treating the value of each impression independently or setting a bid price to each segment of ad volume. However, the bidding for a given ad campaign would repeatedly happen during its life span before the budget runs out. As such, each bid is strategically correlated by the constrained budget and the overall effectiveness of the campaign (e.g., the rewards from generated clicks), which is only observed after the campaign has completed. Thus, it is of great interest to devise an optimal bidding strategy sequentially so that the campaign budget can be dynamically allocated across all the available impressions on the basis of both the immediate and future rewards. In this paper, we formulate the bid decision process as a reinforcement learning problem, where the state space is represented by the auction information and the campaign's real-time parameters, while an action is the bid price to set. By modeling the state transition via auction competition, we build a Markov Decision Process framework for learning the optimal bidding policy to optimize the advertising performance in the dynamic real-time bidding environment. Furthermore, the scalability problem from the large real-world auction volume and campaign budget is well handled by state value approximation using neural networks. The empirical study on two large-scale real-world datasets and the live A/B testing on a commercial platform have demonstrated the superior performance and high efficiency compared to state-of-the-art methods.

In this paper, we deal with the uncertainty of bidding for display advertising. Similar to the financial market trading, real-time bidding (RTB) based display advertising employs an auction mechanism to automate the impression level media buying; and running a campaign is no different than an investment of acquiring new customers in return for obtaining additional converted sales. Thus, how to optimally bid on an ad impression to drive the profit and return-on-investment becomes essential. However, the large randomness of the user behaviors and the cost uncertainty caused by the auction competition may result in a significant risk from the campaign performance estimation. In this paper, we explicitly model the uncertainty of user click-through rate estimation and auction competition to capture the risk. We borrow an idea from finance and derive the value at risk for each ad display opportunity. Our formulation results in two risk-aware bidding strategies that penalize risky ad impressions and focus more on the ones with higher expected return and lower risk. The empirical study on real-world data demonstrates the effectiveness of our proposed risk-aware bidding strategies: yielding profit gains of 15.4% in offline experiments and up to 17.5% in an online A/B test on a commercial RTB platform over the widely applied bidding strategies.

A large number of patients discuss treatments in online health communities (OHCs). One research question of interest to health researchers is whether treatments being discussed in OHCs are eventually used by community members in their real lives. In this paper, we rely on machine learning methods to automatically identify attributions of mentions of treatments from an online autism community. The context of our work is online autism communities, where parents exchange support for the care of their children with autism spectrum disorder. Our methods are able to distinguish discussions of treatments that are associated with patients, caregivers, and others, as well as identify whether a treatment is actually taken. We investigate treatments that are not just discussed but also used by patients according to two types of content analysis, cross-sectional and longitudinal. The treatments identified through our content analysis help create a catalogue of real-world treatments. This study results lay the foundation for future research to compare real-world drug usage with established clinical guidelines.

We address the problem of learning behaviour policies to optimise online metrics from heterogeneous usage data. While online metrics, e.g., click-through rate, can be optimised effectively using exploration data, such data is costly to collect in practice, as it temporarily degrades the user experience. Leveraging related data sources to improve online performance would be extremely valuable, but is not possible using current approaches. We formulate this task as a policy transfer learning problem, and propose a first solution, called collective noise contrastive estimation (collective NCE). NCE is an efficient solution to approximating the gradient of a log-softmax objective. Our approach jointly optimises embeddings of heterogeneous data to transfer knowledge from the source domain to the target domain. We demonstrate the effectiveness of our approach by learning an effective policy for an online radio station jointly from user-generated playlists, and usage data collected in an exploration bucket.

Learning and predicting user responses, such as clicks and conversions, are crucial for many Internet-based businesses including web search, e-commerce, and online advertising. Typically, a user response model is established by optimizing the prediction accuracy, e.g., minimizing the error between the prediction and the ground truth user response. However, in many practical cases, predicting user responses is only part of a rather larger predictive or optimization task, where on one hand, the accuracy of a user response prediction determines the final (expected) utility to be optimized, but on the other hand, its learning may also be influenced from the follow-up stochastic process. It is, thus, of great interest to optimize the entire process as a whole rather than treat them independently or sequentially. In this paper, we take real-time display advertising as an example, where the predicted user's ad click-through rate (CTR) is employed to calculate a bid for an ad impression in the second price auction. We reformulate a common logistic regression CTR model by putting it back into its subsequent bidding context: rather than minimizing the prediction error, the model parameters are learned directly by optimizing campaign profit. The gradient update resulted from our formulations naturally fine-tunes the cases where the market competition is high, leading to a more cost-effective bidding. Our experiments demonstrate that, while maintaining comparable CTR prediction accuracy, our proposed user response learning leads to campaign profit gains as much as 78.2% for offline test and 25.5% for online A/B test over strong baselines.

State-of-the-art item recommendation algorithms, which apply Factorization Machines (FM) as a scoring function and pairwise ranking loss as a trainer (PRFM for short), have been recently investigated for the implicit feedback based context-aware recommendation problem (IFCAR). However, good recommenders particularly emphasize on the accuracy near the top of the ranked list, and typical pairwise loss functions might not match well with such a requirement. In this paper, we demonstrate, both theoretically and empirically, PRFM models usually lead to non-optimal item recommendation results due to such a mismatch. Inspired by the success of LambdaRank, we introduce Lambda Factorization Machines (LambdaFM), which is particularly intended for optimizing ranking performance for IFCAR. We also point out that the original lambda function suffers from the issue of expensive computational complexity in such settings due to a large amount of unobserved feedback. Hence, instead of directly adopting the original lambda strategy, we create three effective lambda surrogates by conducting a theoretical analysis for lambda from the top-N optimization perspective. Further, we prove that the proposed lambda surrogates are generic and applicable to a large set of pairwise ranking loss functions. Experimental results demonstrate LambdaFM significantly outperforms state-of-the-art algorithms on three real-world datasets in terms of four standard ranking measures.

Predicting user responses, such as clicks and conversions, is of great importance and has found its usage inmany Web applications including recommender systems, websearch and online advertising. The data in those applicationsis mostly categorical and contains multiple fields, a typicalrepresentation is to transform it into a high-dimensional sparsebinary feature representation via one-hot encoding. Facing withthe extreme sparsity, traditional models may limit their capacityof mining shallow patterns from the data, i.e. low-order featurecombinations. Deep models like deep neural networks, on theother hand, cannot be directly applied for the high-dimensionalinput because of the huge feature space. In this paper, we proposea Product-based Neural Networks (PNN) with an embeddinglayer to learn a distributed representation of the categorical data, a product layer to capture interactive patterns between interfieldcategories, and further fully connected layers to explorehigh-order feature interactions. Our experimental results on twolarge-scale real-world ad click datasets demonstrate that PNNsconsistently outperform the state-of-the-art models on various metrics.

In real-time display advertising, ad slots are sold per impression via an auction mechanism. For an advertiser, the campaign information is incomplete --- the user responses (e.g, clicks or conversions) and the market price of each ad impression are observed only if the advertiser's bid had won the corresponding ad auction. The predictions, such as bid landscape forecasting, click-through rate (CTR) estimation, and bid optimisation, are all operated in the pre-bid stage with full-volume bid request data. However, the training data is gathered in the post-bid stage with a strong bias towards the winning impressions. A common solution for learning over such censored data is to reweight data instances to correct the discrepancy between training and prediction. However, little study has been done on how to obtain the weights independent of previous bidding strategies and consequently integrate them into the final CTR prediction and bid generation steps. In this paper, we formulate CTR estimation and bid optimisation under such censored auction data. Derived from a survival model, we show that historic bid information is naturally incorporated to produce Bid-aware Gradient Descents (BGD) which controls both the importance and the direction of the gradient to achieve unbiased learning. The empirical study based on two large-scale real-world datasets demonstrates remarkable performance gains from our solution. The learning framework has been deployed on Yahoo!'s real-time bidding platform and provided 2.97% AUC lift for CTR estimation and 9.30% eCPC drop for bid optimisation in an online A/B test.

Real-Time Bidding (RTB) is revolutionising display advertising by facilitating per-impression auctions to buy ad impressions as they are being generated. Being able to use impression-level data, such as user cookies, encourages user behaviour targeting, and hence has significantly improved the effectiveness of ad campaigns. However, a fundamental drawback of RTB is its instability because the bid decision is made per impression and there are enormous fluctuations in campaigns' key performance indicators (KPIs). As such, advertisers face great difficulty in controlling their campaign performance against the associated costs. In this paper, we propose a feedback control mechanism for RTB which helps advertisers dynamically adjust the bids to effectively control the KPIs, e.g., the auction winning ratio and the effective cost per click. We further formulate an optimisation framework to show that the proposed feedback control mechanism also has the ability of optimising campaign performance. By settling the effective cost per click at an optimal reference value, the number of campaign's ad clicks can be maximised with the budget constraint. Our empirical study based on real-world data verifies the effectiveness and robustness of our RTB control system in various situations. The proposed feedback control mechanism has also been deployed on a commercial RTB platform and the online test has shown its success in generating controllable advertising performance.

Business-to-consumer (B2C) emails are usually generated by filling structured user data (e.g.purchase, event) into templates. Extracting structured data from B2C emails allows users to track important information on various devices.

We study and formulate arbitrage in display advertising. Real-Time Bidding (RTB) mimics stock spot exchanges and utilises computers to algorithmically buy display ads per impression via a real-time auction. Despite the new automation, the ad markets are still informationally inefficient due to the heavily fragmented marketplaces. Two display impressions with similar or identical effectiveness (e.g., measured by conversion or click-through rates for a targeted audience) may sell for quite different prices at different market segments or pricing schemes. In this paper, we propose a novel data mining paradigm called Statistical Arbitrage Mining (SAM) focusing on mining and exploiting price discrepancies between two pricing schemes. In essence, our SAMer is a meta-bidder that hedges advertisers' risk between CPA (cost per action)-based campaigns and CPM (cost per mille impressions)-based ad inventories; it statistically assesses the potential profit and cost for an incoming CPM bid request against a portfolio of CPA campaigns based on the estimated conversion rate, bid landscape and other statistics learned from historical data. In SAM, (i) functional optimisation is utilised to seek for optimal bidding to maximise the expected arbitrage net profit, and (ii) a portfolio-based risk management solution is leveraged to reallocate bid volume and budget across the set of campaigns to make a risk and return trade-off. We propose to jointly optimise both components in an EM fashion with high efficiency to help the meta-bidder successfully catch the transient statistical arbitrage opportunities in RTB. Both the offline experiments on a real-world large-scale dataset and online A/B tests on a commercial platform demonstrate the effectiveness of our proposed solution in exploiting arbitrage in various model settings and market environments.

With the increasing accessibility of transactional data in venture finance, venture capital firms (VCs) face great challenges in developing quantitative tools to identify new investment opportunities. Recommendation techniques have the possibility of helping VCs making data-driven investment decisions by providing an automatic screening process of a large number of startups across different domains on the basis of their past investment data. A previous study has shown the potential advantage of using collaborative filtering to catch and predict the VCs' investment behaviours. However, two fundamental challenges in venture finance make conventional recommendation techniques difficult to apply. First, risk factors should be cautiously considered when making investments: for a potential startup, a VC needs to specifically estimate how well this new investment can fit into its holding investment portfolio in such a way that investment risk can be hedged. Second, The investment behaviours are much sparser than conventional recommendation applications and a VC's investments are usually limited to a few industry categories, making it impossible to use a topic-diversification method to hedge the risk. In this paper, we solve the startup recommendation problem from a risk management perspective. We propose 5 risk-aware startup selection and ranking algorithms to catch the VCs' investment behaviours and predict their new investments. Apart from the contribution on the new risk-aware recommendation model, our experiments on the collected CrunchBase dataset show significant performance improvements over strong baselines.

In this paper we study bid optimisation for real-time bidding (RTB) based display advertising. RTB allows advertisers to bid on a display ad impression in real time when it is being generated. It goes beyond contextual advertising by motivating the bidding focused on user data and it is different from the sponsored search auction where the bid price is associated with keywords. For the demand side, a fundamental technical challenge is to automate the bidding process based on the budget, the campaign objective and various information gathered in runtime and in history. In this paper, the programmatic bidding is cast as a functional optimisation problem. Under certain dependency assumptions, we derive simple bidding functions that can be calculated in real time; our finding shows that the optimal bid has a non-linear relationship with the impression level evaluation such as the click-through rate and the conversion rate, which are estimated in real time from the impression level features. This is different from previous work that is mainly focused on a linear bidding function. Our mathematical derivation suggests that optimal bidding strategies should try to bid more impressions rather than focus on a small set of high valued impressions because according to the current RTB market data, compared to the higher evaluated impressions, the lower evaluated ones are more cost effective and the chances of winning them are relatively higher. Aside from the theoretical insights, offline experiments on a real dataset and online experiments on a production RTB system verify the effectiveness of our proposed optimal bidding strategies and the functional optimisation framework.

This paper concerns the task of top-N investment opportunity recommendation in the domain of venture finance. By venture finance, specifically, we are interested in the investment activity of venture capital (VC) firms and their investment partners. We have access to a dataset of recorded venture financings (i.e., investments) by VCs and their investment partners in private US companies. This research was undertaken in partnership with Correlation Ventures, a venture capital firm who are pioneering the use of predictive analytics in order to better inform investment decision making. This paper undertakes a detailed empirical study and data analysis then demonstrates the efficacy of recommender systems in this novel application domain.

In this paper, we study collaborative filtering (CF) in an interactive setting, in which a recommender system continuously recommends items to individual users and receives interactive feedback. Whilst users enjoy sequential recommendations, the recommendation predictions are constantly refined using up-to-date feedback on the recommended items. Bringing the interactive mechanism back to the CF process is fundamental because the ultimate goal for a recommender system is about the discovery of interesting items for individual users and yet users' personal preferences and contexts evolve over time during the interactions with the system. This requires us not to distinguish between the stages of collecting information to construct the user profile and making recommendations, but to seamlessly integrate these stages together during the interactive process, with the goal of maximizing the overall recommendation accuracy throughout the interactions. This mechanism naturally addresses the cold-start problem as any user can immediately receive sequential recommendations without providing ratings beforehand. We formulate the interactive CF with the probabilistic matrix factorization (PMF) framework, and leverage several exploitation-exploration algorithms to select items, including the empirical Thompson sampling and upper confidence bound based algorithms. We conduct our experiment on cold-start users as well as warm-start users with drifting taste. Results show that the proposed methods have significant improvements over several strong baselines for the MovieLens, EachMovie and Netflix datasets.

Personalization techniques have been widely adopted in many recommender systems. However, experiments on real-world datasets show that for some users in certain contexts, personalized recommendations do not necessarily perform better than recommendations that rely purely on popularity. Broadly, this can be interpreted by the fact that the parameters of a personalization model are usually estimated from sparse data; the resulting personalized prediction, despite of its low bias, is often volatile. In this paper, we study the problem further by investigating into the ranking of recommendation lists. From a risk management and portfolio retrieval perspective, there is no difference between the popularity-based and the personalized ranking as both of the recommendation outputs can be represented as the trade-off between expected relevance (reward) and associated uncertainty (risk). Through our analysis, we discover common scenarios and provide a technique to predict whether personalization will fail. Besides the theoretical understanding, our experimental results show that the resulting switch algorithm, which decides whether or not to personalize, outperforms the mainstream recommendation algorithms.

Collaborative filtering techniques rely on aggregated user preference data to make personalized predictions. In many cases, users are reluctant to explicitly express their preferences and many recommender systems have to infer them from implicit user behaviors, such as clicking a link in a webpage or playing a music track. The clicks and the plays are good for indicating the items a user liked (i.e., positive training examples), but the items a user did not like (negative training examples) are not directly observed. Previous approaches either randomly pick negative training samples from unseen items or incorporate some heuristics into the learning model, leading to a biased solution and a prolonged training period. In this paper, we propose to dynamically choose negative training samples from the ranked list produced by the current prediction model and iteratively update our model. The experiments conducted on three large-scale datasets show that our approach not only reduces the training time, but also leads to significant performance gains.

With the ever-growing use of textual biomedical data, domain entity annotation has become very important in biomedicine. Previous works on annotating domain entities from biomedical references suffer from several issues, such as a data flexibility problem, language dependency, and limitations with respect to word sense disambiguation. Meanwhile, the Linked Open Data (LOD) Initiative aims at interlinking data from various open knowledge bases. The numbers of entities and properties describing semantic relationships between entities within the linked data cloud have become very large. In this paper, we propose a knowledge-incentive approach for entity annotation in biomedicine, and present Me Detect, a prototype system that we developed based on this approach. With this approach, we over-come the problems of previous works using LOD-based collective annotation. Finally, we present the results of experiments that verify the effectiveness and efficiency of our approach.

This paper is concerned with the joint allocation of bid price and campaign budget in sponsored search. In this application, an advertiser can create a number of campaigns and set a budget for each of them. In a campaign, he/she can further create several ad groups with bid keywords and bid prices. Data analysis shows that many advertisers are dealing with a very large number of campaigns, bid keywords, and bid prices at the same time, which poses a great challenge to the optimality of their campaign management. As a result, the budgets of some campaigns might be too low to achieve the desired performance goals while those of some other campaigns might be wasted; the bid prices for some keywords may be too low to win competitive auctions while those of some other keywords may be unnecessarily high. In this paper, we propose a novel algorithm to automatically address this issue. In particular, we model the problem as a constrained optimization problem, which maximizes the expected advertiser revenue subject to the constraints of the total budget of the advertiser and the ranges of bid price change. By solving this optimization problem, we can obtain an optimal budget allocation plan as well as an optimal bid price setting. Our simulation results based on the sponsored search log of a commercial search engine have shown that by employing the proposed method, we can effectively improve the performances of the advertisers while at the same time we also see an increase in the revenue of the search engine. In addition, the results indicate that this method is robust to the second-order effects caused by the bid fluctuations from other advertisers.

Digital music has experienced a quite fascinating transformation during the past decades. Thousands of people share or distribute their music collections on the Internet, resulting in an explosive increase of information and more user dependence on automatic recommender systems. Though there are many techniques such as collaborative filtering, most approaches focus mainly on users' global behaviors, neglecting local actions and the specific properties of music. In this paper, we propose a simple and effective local implicit feedback model mining users' local preferences to get better recommendation performance in both rating and ranking prediction. Moreover, we design an efficient training algorithm to speed up the updating procedure, and give a method to find the most appropriate time granularity to assist the performance. We conduct various experiments to evaluate the performance of this model, which show that it outperforms baseline model significantly. Integration with existing temporal models achieves a great improvement compared to the reported best single model for Yahoo! Music.

In recent years, more and more images have been uploaded and published on the Web. Along with text Web pages, images have been becoming important media to place relevant advertisements. Visual contextual advertising, a young research area, refers to finding relevant text advertisements for a target image without any textual information (e.g., tags). There are two existing approaches, advertisement search based on image annotation, and more recently, advertisement matching based on feature translation between images and texts. However, the state of the art fails to achieve satisfactory results due to the fact that recommended advertisements are syntactically matched but semantically mismatched. In this paper, we propose a semantic approach to improving the performance of visual contextual advertising. More specifically, we exploit a large high-quality image knowledge base (ImageNet) and a widely-used text knowledge base (Wikipedia) to build a bridge between target images and advertisements. The image-advertisement match is built by mapping images and advertisements into the respective knowledge bases and then finding semantic matches between the two knowledge bases. The experimental results show that semantic match outperforms syntactic match significantly using test images from Flickr. We also show that our approach gives a large improvement of 16.4% on the precision of the top 10 matches over previous work, with more semantically relevant advertisements recommended.

Recently, recommender systems have fascinated researchers and benefited a variety of people's online activities, enabling users to survive the explosive web information. Traditional collaborative filtering techniques handle the general recommendation well. However, most such approaches usually focus on long term preferences. To discover more short term factors influencing people's decisions, we propose a short term preferences model, implemented with implicit user feedback. We conduct experiments comparing the performances of different short term models, which show that our model outperforms significantly compared to those long term models.

Serendipitous recommendation has benefitted both e-retailers and users. It tends to suggest items which are both unexpected and useful to users. These items are not only profitable to the retailers but also surprisingly suitable to consumers' tastes. However, due to the imbalance in observed data for popular and tail items, existing collaborative filtering methods fail to give satisfactory serendipitous recommendations. To solve this problem, we propose a simple and effective method, called serendipitous personalized ranking. The experimental results demonstrate that our method significantly improves both accuracy and serendipity for top-N recommendation compared to traditional personalized ranking methods in various settings.

Advertising keywords recommendation is an indispensable component for online advertising with the keywords selected from the target Web pages used for contextual advertising or sponsored search. Several ranking-based algorithms have been proposed for recommending advertising keywords. However, for most of them performance is still lacking, especially when dealing with short-text target Web pages, that is, those containing insufficient textual information for ranking. In some cases, short-text Web pages may not even contain enough keywords for selection. A natural alternative is then to recommend relevant keywords not present in the target Web pages. In this article, we propose a novel algorithm for advertising keywords recommendation for short-text Web pages by leveraging the contents of Wikipedia, a user-contributed online encyclopedia. Wikipedia contains numerous entities with related entities on a topic linked to each other. Given a target Web page, we propose to use a content-biased PageRank on the Wikipedia graph to rank the related entities. Furthermore, in order to recommend high-quality advertising keywords, we also add an advertisement-biased factor into our model. With these two biases, advertising keywords that are both relevant to a target Web page and valuable for advertising are recommended. In our experiments, several state-of-the-art approaches for keyword recommendation are compared. The experimental results demonstrate that our proposed approach produces substantial improvement in the precision of the top 20 recommended keywords on short-text Web pages over existing approaches.