With the rapid development of e-commerce, auto-bidding has become a key asset in optimizing advertising performance under diverse advertiser environments. The current approaches focus on reinforcement learning (RL) and generative models. These efforts imitate offline historical behaviors by utilizing a complex structure with expensive hyperparameter tuning. The suboptimal trajectories further exacerbate the difficulty of policy learning. To address these challenges, we proposes QGA, a novel Q-value regularized Generative Auto-bidding method. In QGA, we propose to plug a Q-value regularization with double Q-learning strategy into the Decision Transformer backbone. This design enables joint optimization of policy imitation and action-value maximization, allowing the learned bidding policy to both leverage experience from the dataset and alleviate the adverse impact of the suboptimal trajectories. Furthermore, to safely explore the policy space beyond the data distribution, we propose a Q-value guided dual-exploration mechanism, in which the DT model is conditioned on multiple return-to-go targets and locally perturbed actions. This entire exploration process is dynamically guided by the aforementioned Q-value module, which provides principled evaluation for each candidate action. Experiments on public benchmarks and simulation environments demonstrate that QGA consistently achieves superior or highly competitive results compared to existing alternatives. Notably, in large-scale real-world A/B testing, QGA achieves a 3.27% increase in Ad GMV and a 2.49% improvement in Ad ROI.

Automated bidding is central to modern digital advertising. Early rule-based methods lacked adaptability, while subsequent Reinforcement Learning approaches modeled bidding as a Markov Decision Process but struggled with long-term dependencies. Recent generative models show promise, yet they lack explicit mechanisms to balance exploration and safety, relying solely on action perturbations or trajectory guidance without a safety fallback. This results in inefficient exploration and elevated financial risk for advertising platforms. To address this gap, we propose GUIDE (Generative Auto-Bidding with Unified Modeling and Exploration), a framework that synergistically integrates directed exploration with a safe fallback mechanism. GUIDE employs a Decision Transformer (DT) to jointly model historical bidding actions and environmental state transitions. A Q-value module guides the DT's exploration via regularization constraints, while an Inverse Dynamics Module (IDM) leverages DT-predicted future states to infer robust, behaviorally consistent actions as a safe policy fallback. The Q-value module then adaptively selects the final action between these two options, balancing exploration and safety. Together, these components form an integrated "explore-safeguard-select" pipeline that unifies efficiency and safety. We conduct extensive experiments on public datasets, in simulated auction environments, and through large-scale online deployment on Taobao, a leading Chinese advertising platform. Results show GUIDE consistently outperforms state-of-the-art baselines across all scenarios. In real-world deployment, GUIDE achieves notable gains: +4.10% ad GMV, +1.40% ad clicks, +1.66% ad cost, and +3.52% ad ROI, demonstrating its effectiveness and strong industrial applicability.

Scaling laws have enabled large language models(LLMs) to achieve remarkable performance and strong generalization across diverse language understanding tasks, including few-shot, in-context, and zero-shot learning. While prior studies in large-scale collaborative filtering(CF) have revealed clear relationships between model performance and scaling factors such as data size and model capacity, little attention has been given to how heterogeneous datasets can be synergistically combined for recommender systems(RS). In particular, it remains unclear whether systematically integrating diverse recommendation datasets can yield scaling behaviors analogous to those observed in LLMs, while simultaneously addressing challenges such as cold-start recommendation and cross-domain transfer. In this paper, we present RecCLIP, a multimodal framework that reformulates user--item interactions as visual representations compatible with vision--language models(VLMs). RecCLIP compresses interaction signals and employs prompt-based ranking to enable unified representation across heterogeneous data sources. Extensive experiments reveal consistent power-law scaling trends with respect to data size, and demonstrate that RecCLIP achieves superior performance in both cold-start and cross-domain transfer scenarios. Our findings underscore the importance of data-centric design in recommender systems and provide practical insights into scaling them effectively.The code for replication is available at https://github.com/jinliwei-1/RecCLIP.

Search engines, Recommender systems, and Online advertising are playing fundamental roles in modern web and mobile applications. In these information systems, the most significant component is the ranking system, which selects a list of items likely to interest a user from billions of candidate items. At its core, Deep learning to Rank (DLTR) has become indispensable for building high-performance ranking models, driving significant gains in user engagement and business growth. In this article, firstly, we outline the key problems and challenges in industrial-scale ranking systems. Secondly, we provide a comprehensive review of deep learning models deployed across multiple stages of the industrial ranking pipeline, including matching, pre-ranking, fine-grained ranking, post-ranking, and relevance-ranking. Finally, we explore novel perspectives for future research, such as leveraging Large Language Models (LLMs). The papers discussed in this survey are listed in https://github.com/guyulongcs/Awesome-Deep-Learning-Papers-for-Search-Recommendation-Advertising .

Existing recommendation methods struggle to model users’ multifaceted preferences due to the diversity and volatility of user behavior, as well as the inherent uncertainty and ambiguity of item themes in practical scenarios. Multi-interest recommendation addresses this challenge by explicitly extracting multiple interest representations from users’ historical interactions, enabling fine-grained preference modeling and more accurate recommendations. It has attracted considerable attention in recommendation research. However, current recommendation surveys have either delved into specific recommendation tasks and downstream applications or focused on approaches that model users and items as single representations with cutting-edge techniques, overlooking users’ diverse preferences and the multifaceted aspects of items. In this work, we systematically review the progress, solutions, challenges, and future directions of multi-interest recommendation by answering the following three questions: (1) Why is multi-interest modeling significantly important for recommendation? (2) What aspects are focused on by multi-interest modeling in recommendation? and (3) How can multi-interest modeling be applied, along with the technical details of the representative modules? We hope that this survey establishes a fundamental framework and delivers a preliminary overview for researchers interested in this field and committed to further exploration. The implementation of multi-interest recommendation summarized in this survey is maintained at https://github.com/WHUIR/Multi-Interest-Recommendation-A-Survey .

This is a corrigendum for the article “One Model for All: Large Language Models are Domain-Agnostic Recommendation Systems” published in ACM Trans. Inf. Syst . 43, 5, Article 118 (July 2025), 27 pages.

Generative retrieval presents a promising approach to information retrieval, streamlining both indexing and retrieval processes through end-to-end optimization. This method typically involves assigning a unique identifier to each document, with the retrieval goal being the generation of the correct document identifier in response to a query. Although generative retrieval has demonstrated empirical success in various tasks, designing an effective document identifier remains a challenge. Previous studies have either depended excessively on one-to-one discrete identifiers, leading to increased retrieval latency and loss of semantics in documents or have used retrieval-agnostic dense document identifiers, which can hinder performance. To this end, we propose to integrate the benefits of generative retrieval and dense retrieval using an encoder-decoder-based pre-trained language model. Particularly, the decoder, i.e., the discrete identifier, functions as a coarse retriever, effectively reducing the retrieval space in an end-to-end manner. As a complement, the encoder, i.e., the dense vector, serves as a fine-grained retriever, efficiently and precisely ranking documents in a condensed space. Accordingly, we introduce a three-stage end-to-end learning framework that optimizes identifiers and vectors. Extensive experiments reveal that the proposed method exceeds the current models in terms of effectiveness and time efficiency, across both small and larger corpus sets.

This is an erratum for the article “Learning Discrete Identifiers and Dense Vectors for Generative Retrieval” published in ACM Trans. Inf. Syst. 44, 2, Article 42 (December 2025), 24 pages.

Cross-domain recommendation (CDR) has gained significant attention for its ability to address data sparsity issue. However, most existing CDR methods focus primarily on improving recommendation accuracy while largely overlooking fairness considerations, which can lead to biased outcomes and unfair treatment of different user groups. To solve this critical problem, we investigate whether fairness can be transferred from the source domain to the target domain. Our analysis suggests that fairness can be effectively transferred if the fairness of the source domain is ensured and the distributions of the source and target domains are well aligned. Based on this, we propose the FairCDR, a novel framework that can achieve the knowledge transfer of fairness and user preferences simultaneously. FairCDR owns two phases: single-domain fairness guarantee and inter-domain distribution alignment. In the first phase, we employ an adversarial learning-based recommender (ALR) to disentangle user preferences from sensitive attributes in the source domain. In the second phase, we introduce a new mutual learning-based diffusion model (MLDiff), which engages in mutual learning with ALR to progressively align the distributions of the source and target domains. This improves ALR's adaptability to distribution shifts, ultimately ensuring fairness and recommendation performance in the target domain. Extensive experiments on multiple real-world cross-domain datasets demonstrate that FairCDR surpasses existing strong baselines in both fairness and recommendation quality.

Zero-shot Relation Triplet Extraction (ZeroRTE) aims to extract relation triplets from texts containing unseen relation types. This capability benefits various downstream information retrieval (IR) tasks. The primary challenge lies in enabling models to generalize effectively to unseen relation categories. Existing approaches typically leverage the knowledge embedded in pre-trained language models to accomplish the generalization process. However, these methods focus solely on fitting the training data during training, without specifically improving the model's generalization performance, resulting in limited generalization capability. For this reason, we explore the integration of bi-level optimization (BLO) with pre-trained language models for learning generalized knowledge directly from the training data, and propose a generative meta-learning framework which exploits the 'learning-to-learn' ability of meta-learning to boost the generalization capability of generative models.

Graph-level out-of-distribution (OOD) detection, which attempts to identify OOD graphs originated from an unknown distribution, is a vital building block for safety-critical applications in Web and society. Current approaches concentrate on how to learn better graph representations, but fail to provide any statistically guarantee on detection results, therefore impeding their deployments in the scenario where detection errors would result in serious consequences. To overcome this critical issue, we propose the Conformal Graph-level Out-of-distribution Detection (CGOD), extending the theory of conformal prediction to graph-level OOD detection with a rigorous control over the false positive rate. In CGOD, we develop a new aggregated non-conformity score function based on the proposed adaptive data augmentation. Through the guidance from two designed metrics, i.e., score consistency and representation diversity, our augmentation strategy can generate multiple non-conformity scores, and aggregating these generated non-conformity scores together is robust to the misleading information. Meanwhile, our score function can perceive the subsequent process of conformal inference, enabling the aggregated non-conformity score to be adaptive to different input graphs and deriving a more accurate score estimation. We conduct experiments on multiple real-world datasets with different empirical settings. Extensive results and model analyses demonstrate the superior performance of our approach over several competitive baselines.

The data sparsity problem has been a long-standing obstacle towards achieving better recommendation performance since it is miserable to estimate the user’s interests from limited historical behaviors. The pre-training paradigm, i.e., learning universal knowledge across a wide spectrum of domains, has increasingly become a new de-facto practice in many fields, especially for adaption to new domains. The merit of this superior generalizability renders it a natural choice to tackle the data sparsity problem for various recommendation scenarios. Hence, several efforts mainly follow masked language modeling or simple data augmentation via contrastive learning to build a pre-trained recommendation model. Our recent work (namely Miracle ) suggests that the common treatment utilizing the masked language modeling is not sufficient for pre-training a recommender system, since a user’s intent could be more complex than predicting the next word or item. The encouraging results demonstrate that the multi-interest modeling could significantly push the frontier of recommender system pre-training. Nevertheless, how to accommodate the temporal dynamics of the user interests seems to be underexplored under both single vector representation and multi-interest schemes. In this article, we aim to incorporate sophisticated temporal information modeling with the current advance in this line. More specifically, we extend Miracle by further considering relative position information and two kinds of relative time interval information jointly when performing multi-interest learning. Then, a sequential process for interest refinement is proposed to learn the subtle nuances of how interests change and shift along the timeline, leading to a more precise representation of user interests. Our extensive experiments on multiple real-world datasets validate the effectiveness of the proposed solution, demonstrating a significant improvement over current state-of-the-art models on these benchmarks. The code is available at https://github.com/WHUIR/Horae .

Multimodal recommendation, which utilizes rich multimodal information to learn user preferences, has attracted significant attention. Most works focus on designing powerful encoders for extracting multimodal features, and simply aggregate the learned features together to make prediction. Consequently, they have a limited capacity to learn the inter-modality knowledge including the modality-shared and modality-unique knowledge. In fact, learning the modality-shared knowledge enables us to align cross-modality data for fusing heterogeneous modality features. Learning the modality-unique knowledge is equally important when recommendation tasks only involve a small amount of shared features and the necessary information is contained within specific modality. In this article, we propose Contrastive Modality-Disentangled Learning (CMDL) to overcome this critical limitation. CMDL exactly captures the inter-modality knowledge by achieving modality disentanglement. Specifically, CMDL first disentangles the initial representation into the modality-invariant and modality-specific representations. Afterwards, CMDL introduces a novel manner of contrastive learning to approximate the MI upper bounds for achieving disentanglement regularization. Building upon the proposed regularization, CMDL encourages the modality-invariant and modality-specific representations to capture the modality-shared and modality-unique knowledge respectively and to be statistically independent to each other. Empirically, extensive experiments are conducted on benchmark datasets, demonstrating the superior performance of CMDL compared with strong multimodal recommenders.

Sequential recommendation systems aim to predict users’ next likely interaction based on their history. However, these systems face data sparsity and cold-start problems. Utilizing data from other domains, known as multi-domain methods, is useful for alleviating these problems. However, traditional multi-domain methods rely on meaningless ID-based item representation, which makes it difficult to align items with similar meanings from different domains, yielding sup-optimal knowledge transfer. This article introduces LLM-Rec , a framework that utilizes pre-trained Large Language Models (LLMs) for domain-agnostic recommendation. Specifically, we mix user’s behaviors from multiple domains and concatenate item titles into a sentence, then use LLMs for generating user and item representations. By mixing behaviors across different domains, we can exploit the knowledge encoded in LLMs to bridge the semantic across over multi-domain behaviors, thus obtaining semantically rich representations and improving performance in all domains. Furthermore, we explore the underlying reasons why LLMs are effective and investigate whether LLMs can understand the semantic correlations as the recommendation model, and if advanced techniques like scaling laws in NLP also work in recommendations. We conduct extensive experiments with LLMs ranging from 40 M to 6.7 B to answer the above questions and to verify the effectiveness of LLM-Rec in multi-domain recommendation. The source code is available at https://github.com/WHUIR/LLMRec .

Recently, the paradigm of pre-training and fine-tuning has achieved impressive performance owing to their ability to transfer general knowledge from pre-trained domain to target domain. Meanwhile, graph neural networks (GNNs) have gained prominence in recommender systems. However, there is a lack of unified pre-training and fine-tuning paradigms in graph-based recommendation systems. Applying pre-training and fine-tuning in graph-based recommendation is challenging due to the unique characteristics of recommendation data, including the non-uniform representation, negative transfer effects, and skewed data distributions. To overcome these challenges, we introduce pre-training and prompting recommendation ( ProRec ) , a novel model that synergizes uniform graph pre-training with prompt-tuning for recommendation systems. Specifically, to address the challenge of inconsistent features across different recommendation datasets, ProRec constructs unified input features at the subgraph level and uses a graph auto-encoder for pre-training, laying the foundation for uniform knowledge transfer from the pre-trained domain to the downstream domain. Additionally, ProRec employs prompt-tuning during the fine-tuning phase, which, in a parameter-efficient manner, enhances the generalization of pre-trained knowledge to downstream tasks thereby reducing negative transfer effects. Furthermore, a cross-layer contrastive learning strategy is adopted to eliminate uneven data distribution, promoting more evenly distributed and informative representations. Finally, extensive benchmark comparisons have demonstrated that ProRec outperforms the latest state-of-the-art methods. The source code necessary for replication is available at https://github.com/Code2Q/ProRec .

Recent works demonstrate the effectiveness of multi-modal information for sequential recommendation. However, the computational cost and representation degeneration fail to be focused specifically and addressed adequately in multi-modality recommendation. To this end, we first identify and formalize three properties i.e., diversity, compactness, and consistency from the geometric space and spectrum perspective. Building upon this foundation, we devise tailored loss functions to regularize the above three properties for representation optimization. Theoretical underpinnings and experimental results demonstrate the efficacy of an enhanced item representation in ameliorating degeneration. Furthermore, we propose an efficient and expandable image-centered method, named E2 ImgRec, to mitigate the immense cost of computation. Concretely, we substitute the linear projection operations in the self-attention module and feed-forward network layer with two learnable rescaling vectors or efficient recommendation, then leverage cross-attention for multi-modality information fusion. Extensive experiments on three public datasets illustrate our method outperforms representative ID-based solutions and multi-modal based state-of-the-arts with only up to 39.9% in memory usage and 4.3× acceleration in training time. The code for replication is available at https://github.com/WHUIR/E2ImgRec.

Recommender systems play a pivotal role across practical scenarios, showcasing remarkable capabilities in user preference modeling. However, the centralized learning paradigm predominantly used raises serious privacy concerns. The federated recommender system (FedRS) addresses this by updating models on clients, while a central server orchestrates training without accessing private data. Existing FedRS approaches, however, face unresolved challenges, including non-convex optimization, vulnerability, potential privacy leakage risk, and communication inefficiency. This paper addresses these challenges by reformulating the federated recommendation problem as a convex optimization issue, ensuring convergence to the global optimum. Based on this, we devise a novel method, RFRec, to tackle this optimization problem efficiently. In addition, we propose RFRecF, a highly efficient version that incorporates non-uniform stochastic gradient descent to improve communication efficiency. In user preference modeling, both methods learn local and global models, collaboratively learning users' common and personalized interests under the federated learning setting. Moreover, both methods significantly enhance communication efficiency, robustness, and privacy protection, with theoretical support. Comprehensive evaluations on four benchmark datasets demonstrate RFRec and RFRecF's superior performance compared to diverse baselines. The code is available to ease reproducibility1.

Recently, the item textual information has been exploited with pre-trained language models (PLMs) to enrich the representations of tail items. The underlying idea is to align the hot items and tail items in terms of the external semantic knowledge covered by the PLM. However, it is non-trivial to eliminate the popularity bias by exploiting the textual semantics. One major obstacle is that the model supervision still counts on the sparse yet binary user behaviors. In the preliminary investigation, we discover that text-based recommendations also suffer from the popularity bias.

Online learning to rank (OLTR), which directly optimizes the ranker with interactive user feedback, has gained considerable attention in both academia and industry. However, most current approaches suffer from the inefficiency of heuristic exploration strategies, which can seriously hurt users' experience. Furthermore, the existing OLTR solutions fail to learn from the cost-effective logged data, blocking their usage in the real industrial system. To handle the above issues, we in this paper introduce a new OLTR framework LT2R, namely Learning To online Learning to Rank. LT2R aims to study an efficient parameterized exploration strategy, by which a ranker could converge to the optimal ranking with as few exploration steps as possible. Specifically, we formulate the OLTR task as a typical Markov Decision Process and introduce an online reinforcement learning algorithm with a multi-round cumulative reward to guarantee fast convergence. Moreover, we contribute an offline learning algorithm for LT2R to exploit the knowledge from the historical searching logs, which can provide a fair warm-up model for its industrial deployment. Extensive experiments on both benchmark datasets and Baidu search engine have demonstrated its superiority over state-of-the-art methods.

The logs of the use of a search engine provide sufficient data to train a better ranker. However, it is well known that such implicit feedback reflects biases, and in particular a presentation bias that favors higher-ranked results. Unbiased Learning-to-Rank (ULTR) methods attempt to optimize performance by jointly modeling this bias along with the ranker so that the bias can be removed. Such methods have been shown to provide theoretical soundness, and promise superior performance and low deployment costs. However, existing ULTR methods don't recognize that query-document relevance is a confounder -- it affects both the likelihood of a result being clicked because of relevance and the likelihood of the result being ranked high by the base ranker. Moreover, the performance guarantees of existing ULTR methods assume the use of a weak ranker -- one that does a poor job of ranking documents based on relevance to a query. In practice, of course, commercial search engines use highly tuned rankers, and desire to improve upon them using the implicit judgments in search logs. This results in a significant correlation between position and relevance, which leads existing ULTR methods to overestimate click propensities in highly ranked results, reducing ULTR's effectiveness. This paper is the first to demonstrate the problem of propensity overestimation by ULTR algorithms, based on a causal analysis. We develop a new learning objective based on a backdoor adjustment. In addition, we introduce the Logging-Policy-aware Propensity (LPP) model that can jointly learn LPP and a more accurate ranker. We extensively test our approach on two public benchmark tasks and show that our proposal is effective, practical and significantly outperforms the state of the art.

In the world of E-Commerce, the core task is to understand the personalized preference from various kinds of heterogeneous information, such as textual reviews, item images and historical behaviors. In current systems, these heterogeneous information are mainly exploited to generate better item or user representations. For example, in scenario of visual search, the importance of modeling query image has been widely acknowledged. But, these existing solutions focus on improving the representation quality of the query image, overlooking the personalized visual preference of the user. Note that the visual features affect the user's decision significantly, e.g., the user could be more likely to click the items with her preferred design. Hence, it is fruitful to exploit the visual preference to deliver better capacity for personalization.

The page presentation biases in the information retrieval system, especially on the click behavior, is a well-known challenge that hinders improving ranking models' performance with implicit user feedback. Unbiased Learning to Rank~(ULTR) algorithms are then proposed to learn an unbiased ranking model with biased click data. However, most existing algorithms are specifically designed to mitigate position-related bias, e.g., trust bias, without considering biases induced by other features in search result page presentation(SERP), e.g. attractive bias induced by the multimedia. Unfortunately, those biases widely exist in industrial systems and may lead to an unsatisfactory search experience. Therefore, we introduce a new problem, i.e., whole-page Unbiased Learning to Rank(WP-ULTR), aiming to handle biases induced by whole-page SERP features simultaneously. It presents tremendous challenges: (1) a suitable user behavior model (user behavior hypothesis) can be hard to find; and (2) complex biases cannot be handled by existing algorithms. To address the above challenges, we propose a Bias Agnostic whole-page unbiased Learning to rank algorithm, named BAL, to automatically find the user behavior model with causal discovery and mitigate the biases induced by multiple SERP features with no specific design. Experimental results on a real-world dataset verify the effectiveness of the BAL.

Offline reinforcement learning (RL) makes it possible to train the agents entirely from a previously collected dataset. However, constrained by the quality of the offline dataset, offline RL agents typically have limited performance and cannot be directly deployed. Thus, it is desirable to further finetune the pretrained offline RL agents via online interactions with the environment. Existing offline-to-online RL algorithms suffer from the low sample efficiency issue, due to two inherent challenges, i.e., exploration limitation and distribution shift. To this end, we propose a sample-efficient offline-to-online RL algorithm via Optimistic Exploration and Meta Adaptation (OEMA). Specifically, we first propose an optimistic exploration strategy according to the principle of optimism in the face of uncertainty. This allows agents to sufficiently explore the environment in a stable manner. Moreover, we propose a meta learning based adaptation method, which can reduce the distribution shift and accelerate the offline-to-online adaptation process. We empirically demonstrate that OEMA improves the sample efficiency on D4RL benchmark. Besides, we provide in-depth analyses to verify the effectiveness of both optimistic exploration and meta adaptation.

Recommender systems have been widely adopted in various online personal e-commerce applications for improving user experience. A long-standing challenge in recommender systems is how to provide accurate recommendation to users in cold-start situations where only a few user-item interactions can be observed. Recently, meta learning methods provide a promising solution, and most of them follow a way of parameter initialization where predictions can be fast adapted via multiple gradient descent steps. While these meta-learning recommenders promote model performance, how to derive a fundamental paradigm that enables both flexible approximations of complex user interaction distributions and effective task adaptations of global knowledge still remains a critical yet under-explored problem. To this end, we present the Flow-based Adaptive Neural Process (FANP), a new probabilistic meta-learning model where estimating the preference of each user is governed by an underlying stochastic process. Following an encoder-decoder generative framework, FANP is an effective few-shot function estimator that directly maps limited user interactions to a predictive distribution without complicated gradient updates. Through introducing a conditional normalization flow-based encoder, FANP can get rid of the model bias on latent variables and thereby derive more flexible variational distributions. Meanwhile, we propose a task-adaptive mechanism capturing the relevance of different tasks for improving adaptation ability of global knowledge. The learned task-specific and task-relevant representations are simultaneously exploited to generate the decoder parameters via a novel modulation-augmented hypernetwork. FANP is evaluated on both scenario-specific and user-specific cold-start recommendations on various real-world datasets. Extensive experimental results and detailed model analyses demonstrate that our model yields superior performance compared with multiple state-of-the-art meta-learning recommenders.

User behavior data, such as ratings and clicks, has been widely used to build personalizing models for recommender systems. However, many unflattering factors (e.g., popularity, ranking position, users’ selection) significantly affect the performance of the learned recommendation model. Most existing work on unbiased recommendation addressed these biases from sample granularity (e.g., sample reweighting, data augmentation) or from the perspective of representation learning (e.g., bias-modeling). However, these methods are usually designed for a specific bias, lacking the universal capability to handle complex situations where multiple biases co-exist. Besides, rare work frees itself from laborious and sophisticated debiasing configurations (e.g., propensity scores, imputed values, or user behavior-generating process). Towards this research gap, in this article, we propose a universal G enerative framework for B ias D isentanglement termed as GBD , constantly generating calibration perturbations for the intermediate representations during training to keep them from being affected by the bias. Specifically, a bias-identifier that tries to retrieve the bias-related information from the representations is first introduced. Subsequently, the calibration perturbations are generated to significantly deteriorate the bias-identifier’s performance, making the bias gradually disentangled from the calibrated representations. Therefore, without relying on notorious debiasing configurations, a bias-agnostic model is obtained under the guidance of the bias identifier. We further present its universality by subsuming the representative biases and their mixture under the proposed framework. Finally, extensive experiments on the real-world, synthetic, and semi-synthetic datasets have demonstrated the superiority of the proposed approach against a wide range of recommendation debiasing methods. The code is available at https://github.com/Zhidan-Wang/GBD .

Recommender systems play a pivotal role in navigating users through vast reservoirs of information. However, data sparseness can compromise recommendation accuracy, making it challenging to improve recommendation performance. To address this issue, researchers have explored incorporating multiple data types. Yet, this approach can introduce noise that impairs the recommendations' accuracy. Therefore, it is crucial to denoise the data to enhance recommendation quality. This tutorial highlights the importance of data denoising metrics for improving the accuracy and quality of recommendations. Four groups of data denoising metrics are introduced: feature, item, pattern, and modality level. For each group, various denoising methods are presented. The tutorial emphasizes the significance of selecting the right data denoising methods to enhance recommendation quality. It provides valuable guidance for practitioners and researchers implementing reliable data denoising metrics in recommender systems. Finally, the tutorial proposes open research questions for future studies, making it a valuable resource for the research community.

With the rapid evolution of transformer architectures, researchers are exploring their application in sequential recommender systems (SRSs) and presenting promising performance on SRS tasks compared with former SRS models. However, most existing transformer-based SRS frameworks retain the vanilla attention mechanism, which calculates the attention scores between all item-item pairs. With this setting, redundant item interactions can harm the model performance and consume much computation time and memory. In this paper, we identify the sparse attention phenomenon in transformer-based SRS models and propose Sparse Transformer for sequential Recommendation tasks (STRec) to achieve the efficient computation and improved performance. Specifically, we replace self-attention with cross-attention, making the model concentrate on the most relevant item interactions. To determine these necessary interactions, we design a novel sampling strategy to detect relevant items based on temporal information. Extensive experimental results validate the effectiveness of STRec, which achieves the state-of-the-art accuracy while reducing 54% inference time and 70% memory cost. We also provide massive extended experiments to further investigate the property of our framework.

The pre-training paradigm, i.e., learning universal knowledge across a wide spectrum of domains, has increasingly become a new de-facto practice in many fields, especially for transferring to new domains. The recent progress includes universal pre-training solutions for recommendation. However, we argue that the common treatment utilizing the masked language modeling or simple data augmentation via contrastive learning is not sufficient for pre-training a recommender system, since a user's intent could be more complex than predicting the next word or item. It is more intuitive to go a step further by devising the multi-interest driven pre-training framework for universal user understanding. Nevertheless, incorporating multi-interest modeling in recommender system pre-training is non-trivial due to the dynamic, contextual, and temporary nature of the user interests, particularly when the users are from different domains. The limited effort on this line has greatly rendered it as an open question.

Unbiased Learning to Rank(ULTR), i.e., learning to rank documents with biased user feedback data, is a well-known challenge in information retrieval. Existing methods in unbiased learning to rank typically rely on click modeling or inverse propensity weighting(IPW). Unfortunately, search engines face the issue of a severe long-tail query distribution, which neither click modeling nor IPW handles well. Click modeling usually requires that the same query-document pair appears multiple times for reliable inference, which makes it fall short for tail queries; IPW suffers from high variance since it is highly sensitive to small propensity score values. Therefore, a general debiasing framework that works well under tail queries is sorely needed. To address this problem, we propose a model-based unbiased learning-to-rank framework. Specifically, we develop a general context-aware user simulator to generate pseudo clicks for unobserved ranked lists to train rankers, which addresses the data sparsity problem. In addition, considering the discrepancy between pseudo clicks and actual clicks, we take the observation of a ranked list as the treatment variable and further incorporate inverse propensity weighting with pseudo labels in a doubly robust way. The derived bias and variance indicate that the proposed model-based method is more robust than existing methods. Extensive experiments on benchmark datasets, including simulated datasets and real click logs, demonstrate that the proposed model-based method consistently outperforms state-of-the-art methods in various scenarios. The code is available at https://github.com/rowedenny/MULTR.

Improving user retention with reinforcement learning (RL) has attracted increasing attention due to its significant importance in boosting user engagement. However, training the RL policy from scratch without hurting users’ experience is unavoidable due to the requirement of trial-and-error searches. Furthermore, the offline methods, which aim to optimize the policy without online interactions, suffer from the notorious stability problem in value estimation or unbounded variance in counterfactual policy evaluation. To this end, we propose optimizing user retention with Decision Transformer (DT), which avoids the offline difficulty by translating the RL as an autoregressive problem. However, deploying the DT in recommendation is a non-trivial problem because of the following challenges: (1) deficiency in modeling the numerical reward value; (2) data discrepancy between the policy learning and recommendation generation; (3) unreliable offline performance evaluation. In this work, we, therefore, contribute a series of strategies for tackling the exposed issues. We first articulate an efficient reward prompt by weighted aggregation of meta embeddings for informative reward embedding. Then, we endow a weighted contrastive learning method to solve the discrepancy between training and inference. Furthermore, we design two robust offline metrics to measure user retention. Finally, the significant improvement in the benchmark datasets demonstrates the superiority of the proposed method. The implementation code is available at https://github.com/kesenzhao/DT4Rec.git.

Pre-trained language representation models (PLMs) such as BERT and Enhanced Representation through kNowledge IntEgration (ERNIE) have been integral to achieving recent improvements on various downstream tasks, including information retrieval. However, it is nontrivial to directly utilize these models for the large-scale web search due to the following challenging issues: (1) the prohibitively expensive computations of massive neural PLMs, especially for long texts in the web document, prohibit their deployments in the web search system that demands extremely low latency; (2) the discrepancy between existing task-agnostic pre-training objectives and the ad hoc retrieval scenarios that demand comprehensive relevance modeling is another main barrier for improving the online retrieval and ranking effectiveness; and (3) to create a significant impact on real-world applications, it also calls for practical solutions to seamlessly interweave the resultant PLM and other components into a cooperative system to serve web-scale data. Accordingly, we contribute a series of successfully applied techniques in tackling these exposed issues in this work when deploying the state-of-the-art Chinese pre-trained language model, i.e., ERNIE, in the online search engine system. We first present novel practices to perform expressive PLM-based semantic retrieval with a flexible poly-interaction scheme and cost-efficiently contextualize and rank web documents with a cheap yet powerful Pyramid-ERNIE architecture. We then endow innovative pre-training and fine-tuning paradigms to explicitly incentivize the query-document relevance modeling in PLM-based retrieval and ranking with the large-scale noisy and biased post-click behavioral data. We also introduce a series of effective strategies to seamlessly interwoven the designed PLM-based models with other conventional components into a cooperative system. Extensive offline and online experimental results show that our proposed techniques are crucial to achieving more effective search performance. We also provide a thorough analysis of our methodology and experimental results.

Extracting query-document relevance from the sparse, biased clickthrough log is among the most fundamental tasks in the web search system. Prior art mainly learns a relevance judgment model with semantic features of the query and document and ignores directly counterfactual relevance evaluation from the clicking log. Though the learned semantic matching models can provide relevance signals for tail queries as long as the semantic feature is available. However, such a paradigm lacks the capability to introspectively adjust the biased relevance estimation whenever it conflicts with massive implicit user feedback. The counterfactual evaluation methods, on the contrary, ensure unbiased relevance estimation with sufficient click information. However, they suffer from the sparse or even missing clicks caused by the long-tailed query distribution.

The pre-trained language models (PLMs), such as BERT and ERNIE, have achieved outstanding performance in many natural language understanding tasks. Recently, PLMs-based Information Retrieval models have also been investigated and showed substantially state-of-the-art effectiveness, e.g., MORES, PROP and ColBERT. Moreover, most of the PLMs-based rankers only focus on a single level relevance matching (e.g., character-level), while ignore the other granularity information (e.g., words and phrases), which easily lead to the ambiguity of query understanding and inaccurate matching issues in web search.

Deep pre-trained language models (e.g., BERT) lead to remarkable headway in many Natural Language Processing tasks. Their superior capacity in perceiving textual data is also witnessed in semantic matching tasks (e.g., question answering, web search). Particularly for matching a pair of query and text candidate, the current state-of-the-arts usually rely on the semantic representations produced by BERT, and compute relevance scores with various interaction (i.e., matching) methods. However, they may 1) miss fine-grained phrase-level interaction between the input query and candidate context or 2) lack a thoughtful consideration of both effectiveness and efficiency. Motivated by this, we propose \hyttInteractor, a BERT-based semantic matching model with a flexible contextualized interaction paradigm. It is capable of capturing fine-grained phrase-level information in the interaction, and thus is more effective to be applied for semantic matching tasks. Moreover, we further facilitate \hyttInteractor with a novel partial attention scheme, which significantly reduces the computational cost while maintaining the high effectiveness. We conduct comprehensive experimental evaluations on three datasets. The results show that \hyttInteractor achieves superior effectiveness and efficiency for semantic matching.

Session-based recommendation has recently attracted increasing attention from both industry and academic communities. Previous models mostly focus on designing different models to fit the observed data, which can be quite sparse in real-world scenarios. To alleviate this problem, in this paper, we propose a novel generative session-based recommendation framework. The main building block of our idea is to develop a generator to simulate user sequential behaviors, which are leveraged to train and improve the target sequential recommender model. In order to generate high quality samples, we consider two aspects: (1) the rationality as a sequence of user behaviors, and (2) the informativeness for training the target model. To satisfy these requirements, we design a doubly adversarial network. The first adversarial module aims to make the generated samples conform to the underlying patterns of the real user sequential preference (rationality requirement). The second adversarial module is targeted at widening the model experiences by generating samples which can induce larger model losses (informativeness requirement). In our model, the samples are generated based on a reinforcement learning strategy, where the reward is related with both of the above aspects. In order to stable the training process, we introduce a self-paced regularizer to learn the agent in an easy-to-hard manner. We conduct extensive experiments based on real-world datasets to demonstrate the effectiveness of our model.

As the heart of a search engine, the ranking system plays a crucial role in satisfying users' information demands. More recently, neural rankers fine-tuned from pre-trained language models (PLMs) establish state-of-the-art ranking effectiveness. However, it is nontrivial to directly apply these PLM-based rankers to the large-scale web search system due to the following challenging issues: (1) the prohibitively expensive computations of massive neural PLMs, especially for long texts in the web document, prohibit their deployments in an online ranking system that demands extremely low latency; (2) the discrepancy between existing ranking-agnostic pre-training objectives and the ad-hoc retrieval scenarios that demand comprehensive relevance modeling is another main barrier for improving the online ranking system; (3) a real-world search engine typically involves a committee of ranking components, and thus the compatibility of the individually fine-tuned ranking model is critical for a cooperative ranking system. In this work, we contribute a series of successfully applied techniques in tackling these exposed issues when deploying the state-of-the-art Chinese pre-trained language model, i.e., ERNIE, in the online search engine system. We first articulate a novel practice to cost-efficiently summarize the web document and contextualize the resultant summary content with the query using a cheap yet powerful Pyramid-ERNIE architecture. Then we endow an innovative paradigm to finely exploit the large-scale noisy and biased post-click behavioral data for relevance-oriented pre-training. We also propose a human-anchored fine-tuning strategy tailored for the online ranking system, aiming to stabilize the ranking signals across various online components. Extensive offline and online experimental results show that the proposed techniques significantly boost the search engine's performance.

Post-click conversion, as a strong signal indicating the user preference, is salutary for building recommender systems. However, accurately estimating the post-click conversion rate (CVR) is challenging due to the selection bias, i.e., the observed clicked events usually happen on users' preferred items. Currently, most existing methods utilize counterfactual learning to debias recommender systems. Among them, the doubly robust (DR) estimator has achieved competitive performance by combining the error imputation based (EIB) estimator and the inverse propensity score (IPS) estimator in a doubly robust way. However, inaccurate error imputation may result in its higher variance than the IPS estimator. Worse still, existing methods typically use simple model-agnostic methods to estimate the imputation error, which are not sufficient to approximate the dynamically changing model-correlated target (i.e., the gradient direction of the prediction model). To solve these problems, we first derive the bias and variance of the DR estimator. Based on it, a more robust doubly robust (MRDR) estimator has been proposed to further reduce its variance while retaining its double robustness. Moreover, we propose a novel double learning approach for the MRDR estimator, which can convert the error imputation into the general CVR estimation. Besides, we empirically verify that the proposed learning scheme can further eliminate the high variance problem of the imputation learning. To evaluate its effectiveness, extensive experiments are conducted on a semi-synthetic dataset and two real-world datasets. The results demonstrate the superiority of the proposed approach over the state-of-the-art methods. The code is available at https://github.com/guosyjlu/MRDR-DL.

With the recent advances in Reinforcement Learning (RL), there have been tremendous interests in employing RL for recommender systems. However, directly training and evaluating a new RL-based recommendation algorithm needs to collect users’ real-time feedback in the real system, which is time/effort consuming and could negatively impact users’ experiences. Thus, it calls for a user simulator that can mimic real users’ behaviors to pre-train and evaluate new recommendation algorithms. Simulating users’ behaviors in a dynamic system faces immense challenges – (i) the underlying item distribution is complex, and (ii) historical logs for each user are limited. In this paper, we develop a user simulator based on a Generative Adversarial Network (GAN). To be specific, the generator captures the underlying distribution of users’ historical logs and generates realistic logs that can be considered as augmentations of real logs; while the discriminator not only distinguishes real and fake logs but also predicts users’ behaviors. The experimental results based on benchmark datasets demonstrate the effectiveness of the proposed simulator.

Recommender Systems have been playing essential roles in e-commerce portals. Existing recommendation algorithms usually learn the ranking scores of items by optimizing a single task (e.g. Click-through rate prediction) based on users' historical click sequences, but they generally pay few attention to simultaneously modeling users' multiple types of behaviors or jointly optimize multiple objectives (e.g. both Click-through rate and Conversion rate), which are both vital for e-commerce sites. In this paper, we argue that it is crucial to formulate users' different interests based on multiple types of behaviors and perform multi-task learning for significant improvement in multiple objectives simultaneously. We propose Deep Multifaceted Transformers (DMT), a novel framework that can model users' multiple types of behavior sequences simultaneously with multiple Transformers. It utilizes Multi-gate Mixture-of-Experts to optimize multiple objectives. Besides, it exploits unbiased learning to reduce the selection bias in the training data. Experiments on JD real production dataset demonstrate the effectiveness of DMT, which significantly outperforms state-of-art methods. DMT has been successfully deployed to serve the main traffic in the commercial Recommender System in JD.com. To facilitate future research, we release the codes and datasets at https://github.com/guyulongcs/CIKM2020_DMT.

Distant supervision provides a means to create a large number of weakly labeled data at low cost for relation classification. However, the resulting labeled instances are very noisy, containing data with wrong labels. Many approaches have been proposed to select a subset of reliable instances for neural model training, but they still suffer from noisy labeling problem or underutilization of the weakly-labeled data. To better select more reliable training instances, we introduce a small amount of manually labeled data as reference to guide the selection process. In this paper, we propose a meta-learning based approach, which learns to reweight noisy training data under the guidance of reference data. As the clean reference data is usually very small, we propose to augment it by dynamically distilling the most reliable elite instances from the noisy data. Experiments on several datasets demonstrate that the reference data can effectively guide the selection of training data, and our augmented approach consistently improves the performance of relation classification comparing to the existing state-of-the-art methods.

With the recent prevalence of Reinforcement Learning (RL), there have been tremendous interests in developing RL-based recommender systems. In practical recommendation sessions, users will sequentially access multiple scenarios, such as the entrance pages and the item detail pages, and each scenario has its specific characteristics. However, the majority of existing RL-based recommender systems focus on optimizing one strategy for all scenarios or separately optimizing each strategy, which could lead to sub-optimal overall performance. In this paper, we study the recommendation problem with multiple (consecutive) scenarios, i.e., whole-chain recommendations. We propose a multi-agent RL-based approach (DeepChain), which can capture the sequential correlation among different scenarios and jointly optimize multiple recommendation strategies. To be specific, all recommender agents (RAs) share the same memory of users' historical behaviors, and they work collaboratively to maximize the overall reward of a session. Note that optimizing multiple recommendation strategies jointly faces two challenges in the existing model-free RL model - (i) it requires huge amounts of user behavior data, and (ii) the distribution of reward (users' feedback) are extremely unbalanced. In this paper, we introduce model-based RL techniques to reduce the training data requirement and execute more accurate strategy updates. The experimental results based on a real e-commerce platform demonstrate the effectiveness of the proposed framework.

In this paper, we study collaborative filtering in an interactive setting, in which the recommender agents iterate between making recommendations and updating the user profile based on the interactive feedback. The most challenging problem in this scenario is how to suggest items when the user profile has not been well established, \ie recommend for cold-start users or warm-start users with taste drifting. Existing approaches either rely on overly pessimistic linear exploration strategy or adopt meta-learning based algorithms in a full exploitation way. In this work, to quickly catch up with the user's interests, we proposed to represent the exploration policy with a neural network and directly learn it from the feedback data. Specifically, the exploration policy is encoded in the weights of multi-channel stacked self-attention neural networks and trained with efficient Q-learning by maximizing users' overall satisfaction in the recommender systems. The key insight is that the satisfied recommendations triggered by the exploration recommendation can be viewed as the exploration bonus (delayed reward) for its contribution on improving the quality of the user profile. Therefore, the proposed exploration policy, to balance between learning the user profile and making accurate recommendations, can be directly optimized by maximizing users' long-term satisfaction with reinforcement learning. Extensive experiments and analysis conducted on three benchmark collaborative filtering datasets have demonstrated the advantage of our method over state-of-the-art methods.

Applying reinforcement learning (RL) in recommender systems is attractive but costly due to the constraint of the interaction with real customers, where performing online policy learning through interacting with real customers usually harms customer experiences. A practical alternative is to build a recommender agent offline from logged data, whereas directly using logged data offline leads to the problem of selection bias between logging policy and the recommendation policy. The existing direct offline learning algorithms, such as Monte Carlo methods and temporal difference methods are either computationally expensive or unstable on convergence. To address these issues, we propose Pseudo Dyna-Q (PDQ). In PDQ, instead of interacting with real customers, we resort to a customer simulator, referred to as the World Model, which is designed to simulate the environment and handle the selection bias of logged data. During policy improvement, the World Model is constantly updated and optimized adaptively, according to the current recommendation policy. This way, the proposed PDQ not only avoids the instability of convergence and high computation cost of existing approaches but also provides unlimited interactions without involving real customers. Moreover, a proved upper bound of empirical error of reward function guarantees that the learned offline policy has lower bias and variance. Extensive experiments demonstrated the advantages of PDQ on two real-world datasets against state-of-the-arts methods.

Recommender systems play a crucial role in our daily lives. Feed streaming mechanism has been widely used in the recommender system, especially on the mobile Apps. The feed streaming setting provides users the interactive manner of recommendation in never-ending feeds. In such a manner, a good recommender system should pay more attention to user stickiness, which is far beyond classical instant metrics and typically measured by long-term user engagement. Directly optimizing long-term user engagement is a non-trivial problem, as the learning target is usually not available for conventional supervised learning methods. Though reinforcement learning~(RL) naturally fits the problem of maximizing the long term rewards, applying RL to optimize long-term user engagement is still facing challenges: user behaviors are versatile to model, which typically consists of both instant feedback (eg. clicks) and delayed feedback (eg. dwell time, revisit); in addition, performing effective off-policy learning is still immature, especially when combining bootstrapping and function approximation. To address these issues, in this work, we introduce a RL framework --- FeedRec to optimize the long-term user engagement. FeedRec includes two components: 1)~a Q-Network which designed in hierarchical LSTM takes charge of modeling complex user behaviors, and 2)~a S-Network, which simulates the environment, assists the Q-Network and voids the instability of convergence in policy learning. Extensive experiments on synthetic data and a real-world large scale data show that FeedRec effectively optimizes the long-term user engagement and outperforms state-of-the-arts.

In e-commerce, users' demands are not only conditioned by their profile and preferences, but also by their recent purchases that may generate new demands, as well as periodical demands that depend on purchases made some time ago. We call them respectively short-term demands and long-term demands. In this paper, we propose a novel self-attentive Continuous-Time Recommendation model (CTRec) for capturing the evolving demands of users over time. For modeling such time-sensitive demands, a Demand-aware Hawkes Process (DHP) framework is designed in CTRec to learn from the discrete purchase records of users. More specifically, a convolutional neural network is utilized to capture the short-term demands; and a self-attention mechanism is employed to capture the periodical purchase cycles of long-term demands. All types of demands are fused in DHP to make final continuous-time recommendations. We conduct extensive experiments on four real-world commercial datasets to demonstrate that CTRec is effective for general sequential recommendation problems, including next-item and next-session/basket recommendations. We observe in particular that CTRec is capable of learning the purchase cycles of products and estimating the purchase time of a product given a user.

The rapid spread of mobile internet and location-acquisition technologies have led to the increasing popularity of Location-Based Social Networks(LBSNs). Users in LBSNs can share their life by checking in at various venues at any time. In LBSNs, identifying home locations of users is significant for effective location-based services like personalized search, targeted advertisement, local recommendation and so on. In this paper, we propose a Home Location Global Positioning System called HLGPS to tackle with the home location identification problem in LBSNs. Firstly, HLGPS uses an influence model named as IME to model edges in LBSNs. Then HLGPS uses a global iteration algorithm based on IME model to position home location of users so that the joint probability of generating all the edges in LBSNs is maximum. Extensive experiments on a large real-world LBSN dataset demonstrate that HLGPS significantly outperforms state-of-the-art methods by 14.7%.

Enormous efforts of human volunteers have made Wikipedia become a treasure of textual knowledge. Relation extraction that aims at extracting structured knowledge in the unstructured texts in Wikipedia is an appealing but quite challenging problem because it's hard for machines to understand plain texts. Existing methods are not effective enough because they understand relation types in textual level without exploiting knowledge behind plain texts. In this paper, we propose a novel framework called Athena 2.0 leveraging Semantic Patterns which are patterns that can understand relation types in semantic level to solve this problem. Extensive experiments show that Athena 2.0 significantly outperforms existing methods.

Event-based Social Networks(EBSNs) which combine online interactions and offline events among users have experienced increased popularity and rapid growth recently. In EBSNs, event recommendation is significant for users due to the extremely large amount of events. However, the event recommendation problem is rather challenging because it faces a serious cold-start problem: Events have short life time and new events are registered by only a few users. What's more, there are only implicit feedback information. Existing approaches like collaborative filtering methods are not suitable for this scenario. In this paper, we propose a Context Aware Matrix Factorization model called AlphaMF to tackle with the problem. Specifically, AlphaMF is a unified model that combines the Matrix Factorization model which models implicit feedbacks with the Linear contextual features model which models explicit contextual features. Extensive experiments on a large real-world EBSN dataset demonstrate that the AlphaMF model significantly outperforms state-of-the-art methods by 11%.