Enabling robots to grasp disorganized cloth for efficient storage is valuable in robot-assisted room organization. Diverse deformations of cloth and the stacking of multiple items limit grasping-pose estimation that relies on annotations. This necessitates segmenting each cloth item in an unsupervised manner before estimating the grasping position. However, existing segmentation methods primarily focus on improving metrics such as Intersection-over-Union and Pixel Accuracy, which cannot effectively measure the segmentation errors of the cloth area and thus lead to failure grasping position estimation. To address this challenge, we use False Discovery Rate (FDR) as a novel measure of segmentation errors and analyze its impact on grasping success. Our preliminary study reveals a negative correlation between segmentation FDR and grasping success rate, highlighting the need for more reliable segmentation in cluttered cloth scenarios. Therefore, we propose an unsupervised cloth segmentation network based on feature distance-weighted constraints, designed to reduce the false discovery rate in cloth area perception without requiring expensive pixel-level manual annotations. Additionally, to estimate the grasping position on the perceived cloth area, we introduce a strategy based on cloth surface wrinkle analysis, which operates without the need for annotations or training. By integrating the proposed segmentation network and grasping strategy, we develop a robotic system capable of sequentially grasping cluttered cloth from a table. Extensive real-world robotic experiments demonstrate the effectiveness of our approach, outperforming multiple baseline methods in segmentation FDR and grasping success rate.

Link prediction is a cornerstone of the Web ecosystem, powering applications from recommendation and search to knowledge graph completion and collaboration forecasting. However, large-scale networks present unique challenges: they contain hundreds of thousands of nodes and edges with heterogeneous and overlapping community structures that evolve over time. Existing approaches face notable limitations: traditional graph neural networks struggle to capture global structural dependencies, while recent graph transformers achieve strong performance but incur quadratic complexity and lack interpretable latent structure. We propose TGSBM (Transformer-Guided Stochastic Block Model), a framework that integrates the principled generative structure of Overlapping Stochastic Block Models with the representational power of sparse Graph Transformers. TGSBM comprises three main components: (i) expander-augmented sparse attention that enables near-linear complexity and efficient global mixing, (ii) a neural variational encoder that infers structured posteriors over community memberships and strengths, and (iii) a neural edge decoder that reconstructs links via OSBM's generative process, preserving interpretability. Experiments across diverse benchmarks demonstrate competitive performance (mean rank 1.6 under HeaRT protocol), superior scalability (up to 6× faster training), and interpretable community structures. These results position TGSBM as a practical approach that strikes a balance between accuracy, efficiency, and transparency for large-scale link prediction.

Recentworks have shown the potential of diffusion models in computer vision and natural language processing. Apart from the classical supervised learning fields, diffusion models have also shown strong competitiveness in reinforcement learning (RL) by formulating decision-making as sequential generation. However, incorporating temporal information of sequential data and utilizing it to guide diffusion models to perform better generation is still an open challenge. In this paper, we take one step forward to investigate controllable generation with temporal conditions that are refined from temporal information. We observe the importance of temporal conditions in sequential generation in sufficient scenarios and provide a comprehensive discussion and comparison of different temporal conditions. Based on the observations, we propose an effective temporally-conditional diffusion model coined Temporally-Composable Diffuser (TCD), which extracts temporal information from interaction sequences and explicitly guides generation with temporal conditions. Specifically, we separate the sequences into three parts according to time expansion and identify historical, immediate, and prospective conditions accordingly. Each condition preserves non-overlapping temporal information of sequences, enabling more controllable generation when we jointly use them to guide the diffuser. Finally, we conduct extensive experiments and analysis to reveal the favorable applicability of TCD in offline RL tasks, where our method reaches or matches the best performance compared with prior SOTA baselines.

In real-world decision-making tasks, it is critical for reinforcement learning (RL) methods to be both stable and robust. Maximum entropy RL methods typically generate a robust policy with entropy augmented reward. While incorporating entropy into the reward offers the benefit of exploration, it presents limited universal applicability and persistent convergence difficulties, such as suboptimal policy stabilization and unstable $Q$ value update. From optimization, we define these two issues as tremulous policy and spiky Q-function, investigating their underlying causes and relationships. Analysis with this, the maximum entropy principle leads to a spiky $Q$ -function update, which ultimately results in a tremulous policy. We thus introduce a beta-symmetric Kullback-Leibler (KL) divergence objective to mitigate such issues under the maximum entropy framework. With this objective function, the tremulous nature of the policy could be controlled with a large beta value. The spiky $Q$ -function could be avoided by annealing the entropy in the target $Q$ value, as the beta-symmetric KL divergence is an upper bound of the original reverse KL divergence. Theoretically, we prove that minimizing our new objective function results in a new policy that presents an improvement in the $Q$ value. Guaranteed by these results, we ultimately derive the optimal policy by iteratively updating the $Q$ value and policy, and we call this method max-entropy stable optimization (MeSO). Experimental results on the Mujoco and Roboschool platforms demonstrate that our algorithm maintains stability while offering better flexibility and overall performance.

Offline reinforcement learning (RL) provides a promising solution to learning an agent fully relying on a data-driven paradigm. However, constrained by the limited quality of the offline dataset, its performance is often suboptimal. Therefore, it is desired to further finetune the agent via extra online interactions before deployment. Unfortunately, offline-to-online RL can be challenging due to two main challenges: constrained exploratory behavior and state-action distribution shift. In view of this, we propose a simple unified uncertainty-guided (SUNG) framework, which naturally unifies the solution to both challenges with the tool of uncertainty. Specifically, SUNG quantifies uncertainty via a variational autoencoder (VAE)-based state-action visitation density estimator. To facilitate efficient exploration, SUNG presents a practical optimistic exploration strategy to select informative actions with both high value and high uncertainty. Moreover, SUNG develops an adaptive exploitation method by applying conservative offline RL objectives to high-uncertainty samples and standard online RL objectives to low-uncertainty samples to smoothly bridge offline and online stages. SUNG achieves state-of-the-art online finetuning performance when combined with different offline RL methods, across various environments and datasets in the D4RL benchmark. Codes are made publicly available in https://github.com/guosyjlu/ACBR.

Graph neural networks (GNNs) have shown significant success in learning graph representations. However, recent studies reveal that GNNs often fail to outperform simple MLPs on heterophilous graph tasks, where connected nodes may differ in features or labels, challenging the homophily assumption. Existing methods addressing this issue often overlook the importance of information granularity and rarely consider implicit relationships between distant nodes. To overcome these limitations, we propose the Granular and Implicit Graph Network (GRAIN), a novel GNN model specifically designed for heterophilous graphs. GRAIN enhances node embeddings by aggregating multi-view information at various granularity levels and incorporating implicit data from distant, non-neighboring nodes. This approach effectively integrates local and global information, resulting in smoother, more accurate node representations. We also introduce an adaptive graph information aggregator that efficiently combines multi-granularity and implicit data, significantly improving node representation quality, as shown by experiments on 13 datasets covering varying homophily and heterophily. GRAIN consistently outperforms 12 state-of-the-art models, excelling on both homophilous and heterophilous graphs.

Traffic congestion is becoming an increasingly prominent problem, and intelligent traffic signal control methods can effectively alleviate it. Recently, there has been a growing trend of applying reinforcement learning to traffic signal control for adaptive signal scheduling. However, most existing methods focus on improving traffic performance while neglecting the issue of scheduling fairness, resulting in long waiting time for some vehicles. Some works attempt to address fairness issues but often sacrifice transport performance. Furthermore, existing methods overlook the challenge of sample efficiency, especially when dealing with diversity-limited traffic data. Therefore, we propose a Fairess-aware and sample-Efficient traffic signal control method called FELight. Specifically, we first design a novel fairness metric and integrate it into decision process to penalize cases with high latency by setting a threshold for activating the fairness mechanism. Theoretical comparison with other fairness works proves why and when our fairness could bring advantages. Moreover, counterfactual data augmentation is employed to enrich interaction data, enhancing the sample efficiency of FELight. Self-supervised state representation is introduced to extract informative features from raw states, further improving sample efficiency. Experiments on real traffic datasets demonstrate that FELight provides relatively fairer traffic signal control without compromising performance compared to state-of-the-art approaches.

Accurately predicting user watch-time is crucial for enhancing user stickiness and retention in video recommendation systems. Existing watch-time prediction approaches typically involve transformations of watch-time labels for prediction and subsequent reversal, ignoring both the natural distribution properties of label and the instance representation confusion that results in inaccurate predictions. In this paper, we propose ProWTP, a two-stage method combining prototype learning and optimal transport for watch-time regression prediction, suitable for any deep recommendation model. Specifically, we observe that the watch-ratio (the ratio of watch-time to video duration) within the same duration bucket exhibits a multimodal distribution. To facilitate incorporation into models, we use a hierarchical vector quantised variational autoencoder (HVQ-VAE) to convert the continuous label distribution into a high-dimensional discrete distribution, serving as credible prototypes for calibrations. Based on this, ProWTP views the alignment between prototypes and instance representations as a Semi-relaxed Unbalanced Optimal Transport (SUOT) problem, where the marginal constraints of prototypes are relaxed. And the corresponding optimization problem is reformulated as a weighted Lasso problem for solution. Moreover, ProWTP introduces the assignment and compactness losses to encourage instances to cluster closely around their respective prototypes, thereby enhancing the prototype-level distinguishability. Finally, we conducted extensive offline experiments on two industrial datasets, demonstrating our consistent superiority in real-world application.

In this work, we explore the potential of large language models (LLMs) for generating functional test scripts, which necessitates understanding the dynamically evolving code structure of the target software. To achieve this, we propose a case-based reasoning (CBR) system utilizing a 4R cycle (i.e., retrieve, reuse, revise, and retain), which maintains and leverages a case bank of test intent descriptions and corresponding test scripts to facilitate LLMs for test script generation. To improve user experience further, we introduce Re4, an optimization method for the CBR system, comprising reranking-based retrieval finetuning and reinforced reuse finetuning. Specifically, we first identify positive examples with high semantic and script similarity, providing reliable pseudo-labels for finetuning the retriever model without costly labeling. Then, we apply supervised finetuning, followed by a reinforcement learning finetuning stage, to align LLMs with our production scenarios, ensuring the faithful reuse of retrieved cases. Extensive experimental results on two product development units from Huawei Datacom demonstrate the superiority of the proposed CBR+Re4. Notably, we also show that the proposed Re4 method can help alleviate the repetitive generation issues with LLMs.

In video recommendation systems, accurately predicting watch time is crucial for enhancing user engagement and retention. Traditional methods typically apply label transformations or mitigate duration bias to improve performance but overlook that erroneous instance representations are the primary cause of significant prediction errors. Moreover, these approaches predominantly rely on point perdition, limiting their robustness. To address these challenges, we propose LDA, a novel prediction paradigm that optimizes instance representations by explicitly leveraging label distributions as anchors within the model, enabling more accurate and robust predictions. Our analysis reveals that watch ratio across different duration groups exhibit distinct multi-peak distributions, reflecting the strong aggregation of user behavior. Based on this finding, we employ Vector Quantized Variational Auto-encoder (VQ-VAE) to convert the continuous watch ratio distribution into representative anchors that capture these multi-peak characteristics within each duration group. Subsequently, we project both instance representations and anchors into a common space and utilize Optimal Transport (OT) to generate pseudo-labels aligned with the anchor distribution, allowing instances to obtain structured coordinates within this space during training. Finally, we derive optimized instance representations for watch time prediction by aggregating anchor vectors through weighted integration. Extensive offline experiments on two datasets and large-scale online A/B testing on a short-video platform with over 300 million DAUs demonstrate the consistent superiority of LDA in watch time prediction.

Long-horizon planning in robotic manipulation tasks requires translating underspecified, symbolic goals into executable control programs satisfying spatial, temporal, and physical constraints. However, language model-based planners often struggle with long-horizon task decomposition, robust constraint satisfaction, and adaptive failure recovery. We introduce InstructFlow, a multi-agent framework that establishes a symbolic, feedback-driven flow of information for code generation in robotic manipulation tasks. InstructFlow employs a InstructFlow Planner to construct and traverse a hierarchical instruction graph that decomposes goals into semantically meaningful subtasks, while a Code Generator generates executable code snippets conditioned on this graph. Crucially, when execution failures occur, a Constraint Generator analyzes feedback and induces symbolic constraints, which are propagated back into the instruction graph to guide targeted code refinement without regenerating from scratch. This dynamic, graph-guided flow enables structured, interpretable, and failure-resilient planning, significantly improving task success rates and robustness across diverse manipulation benchmarks, especially in constraint-sensitive and long-horizon scenarios.

Continual offline reinforcement learning (CORL) has shown impressive ability in diffusion-based continual learning systems by modeling the joint distributions of trajectories. However, most research only focuses on limited continual task settings where the tasks have the same observation and action space, which deviates from the realistic demands of training agents in various environments. In view of this, we propose Vector-Quantized Continual Diffuser, named VQ-CD, to break the barrier of different spaces between various tasks. Specifically, our method contains two complementary sections, where the quantization spaces alignment provides a unified basis for the selective weights activation. In the quantized spaces alignment, we leverage vector quantization to align the different state and action spaces of various tasks, facilitating continual training in the same space. Then, we propose to leverage a unified diffusion model attached by the inverse dynamic model to master all tasks by selectively activating different weights according to the task-related sparse masks. Finally, we conduct extensive experiments on 15 continual learning (CL) tasks, including conventional CL task settings (identical state and action spaces) and general CL task settings (various state and action spaces). Compared with 17 baselines, our method reaches the SOTA performance.

Conditional decision generation with diffusion models has shown powerful competitiveness in reinforcement learning (RL). Recent studies reveal the relation between energy-function-guidance diffusion models and constrained RL problems. The main challenge lies in estimating the intermediate energy, which is intractable due to the log-expectation formulation during the generation process. To address this issue, we propose the Analytic Energy-guided Policy Optimization (AEPO). Specifically, we first provide a theoretical analysis and the closed-form solution of the intermediate guidance when the diffusion model obeys the conditional Gaussian transformation. Then, we analyze the posterior Gaussian distribution in the log-expectation formulation and obtain the target estimation of the log-expectation under mild assumptions. Finally, we train an intermediate energy neural network to approach the target estimation of log-expectation formulation. We apply our method in 30+ offline RL tasks to demonstrate the effectiveness of our method. Extensive experiments illustrate that our method surpasses numerous representative baselines in D4RL offline reinforcement learning benchmarks.

Out-of-distribution (OOD) generalization is critical for applying reinforcement learning algorithms to real-world applications. To address the OOD problem, recent works focus on learning an OOD adaptation policy by capturing the causal factors affecting the environmental dynamics. However, these works recover the causal factors with only an entangled or binary form, resulting in a limited generalization of the policy that requires extra data from the testing environments. To break this limitation, we propose generalizable causal reinforcement learning (GCRL) to learn a disentangled representation of causal factors, on the basis of which we learn a policy that achieves the OOD generalization without extra training. For capturing the causal factors, GCRL deploys a weakly supervised signal with a two-stage constraint to ensure that all factors can be disentangled. Then, to achieve the OOD generalization through causal factors, we establish the dependence of actions on the learned representation and optimize the policy model across multiple environments. Experimental results show that the established dependence recovers the correct relationship between causal factors and actions when the learned policy could address the target tasks in training environments. Benefiting from the recovered relationship, GCRL achieves the OOD generalization on eight benchmarks from Causal World and Mujoco. Moreover, the policy learned by our model is more explainable and can be controlled to generate semantic actions by intervening in the representation of causal factors.

The cooperation in mixed cooperative-competitive tasks has drawn significant attention in multi-agent deep reinforcement learning. Agents need to cooperate with their teammates while competing against their opponents. However, most existing works treat the cooperative agents and competitive agents equally as they perform the same operation on all the agents. As a result, without distinguishing between cooperative and competitive agents, they may suffer from information disorder in learning an optimally cooperative policy and struggle to decide on the next step action. To address the above issues, we decompose the final Q-value into a weighted combination of three parts: the Q-values of the cooperative agents, the competitive group, and the current agent. A theoretical proof of the correctness of the decomposition is provided. With this decomposition, we are able to consider cooperative and competitive agents separately. Accordingly, we propose a multi-agent actor-critic algorithm called actor-double-attention-critic (ADAC) under centralized training and decentralized execution according to the decomposition. In ADAC, networks with group-specific attention and an attentional weighting network are specially designed. With the designed double-attention structure, ADAC can capture the distributions from different agents and improve cooperation performance. Extensive experiments are conducted in three scenarios with nine settings against six representative methods. The results demonstrate the superiority of the proposed ADAC model against state-of-the-art methods in various mixed cooperative-competitive tasks. The code is available at https://github.com/CrazyBayes/ADAC

Graph Neural Networks (GNNs) have been gaining more attention due to their excellent performance in modeling various graph-structured data. However, most of the current GNNs only consider fixed-neighbor discrete message-passing, disregarding the importance of the local structure of different nodes and the implicit information between nodes for smoothing features. Previous approaches either focus on adaptive selection for aggregation structures or treat discrete graph convolution as a continuous diffusion process, but none of them comprehensively considered the above issues, significantly limiting the model's performance. To this end, we present a novel approach called Flexible Diffusion Convolution (Flexi-DC), which exploits the neighborhood information of nodes to set a particular continuous diffusion for each node to smooth features. Specifically, Flexi-DC first extracts the local structure knowledge based on the degrees of nodes in the graph data and then injects it into the diffusion convolution module to smooth features. Additionally, we utilize the extracted knowledge to smooth labels. Flexi-DC is an efficient framework that can significantly improve the performance of most GNN architectures. Experimental results demonstrate that Flexi-DC outperforms their vanilla implementations by an average accuracy of 13.24% (GCN), 16.37% (JKNet), and 11.98% (ARMA) on nine graph datasets with different homophily ratios.

Graph neural networks (GNNs) have garnered significant attention for their competitive performance on graph-structured data. However, many existing methods are commonly constrained by the homophily assumption, making them overly reliant on the uniform neighbor propagation, which limits their ability to generalize to heterophilous graphs. Although some approaches extend aggregation to multi-hop neighbors, adapting neighborhood sizes on a per-node basis remains a significant challenge. In view of this, we propose an Evolutionary Graph Neural Network (EGNN) with adaptive structure-level aggregation and label smoothing, offering a novel solution to the aforementioned drawback. The core innovation of EGNN lies in assigning each node apersonalizedneighborhood structure utilizingbehavior-levelcrossover and mutation. Specifically, we first adaptively search for the optimal structure-level neighborhoods for nodes within the solution space, leveraging the exploratory capabilities of evolutionary computation. This approach enhances the exchange of information between the target node and surrounding nodes, achieving a smooth vector representation. Subsequently, we adopt the optimal structure obtained through evolutionary search to perform label smoothing, further boosting the robustness of the framework. We conduct experiments on nine real-world networks with different homophily ratios, where outstanding performance demonstrates that the ability of EGNN can match or surpass SOTA baselines.

Artificial neural networks, especially recent diffusion-based models, have shown remarkable superiority in gaming, control, and QA systems, where the training tasks' datasets are usually static. However, in real-world applications, such as robotic control of reinforcement learning (RL), the tasks are changing, and new tasks arise in a sequential order. This situation poses the new challenge of plasticity-stability tradeoff for training an agent who can adapt to task changes and retain acquired knowledge. In view of this, we propose a rehearsal-based continual diffusion model, called continual diffuser (CoD), to endow the diffuser with the capabilities of quick adaptation (plasticity) and lasting retention (stability). Specifically, we first construct an offline benchmark that contains 90 tasks from multiple domains. Then, we train the CoD on each task with sequential modeling and conditional generation for making decisions. Next, we preserve a small portion of previous datasets as the rehearsal buffer and replay it to retain the acquired knowledge. Extensive experiments on a series of tasks show that CoD can achieve a promising plasticity-stability tradeoff and outperform existing diffusion-based methods and other representative baselines on most tasks. The source code is available at https://github.com/JF-Hu/Continual_Diffuser.

Multiagent policy gradients (MAPGs), an essential branch of reinforcement learning (RL), have made great progress in both industry and academia. However, existing models do not pay attention to the inadequate training of individual policies, thus limiting the overall performance. We verify the existence of imbalanced training in multiagent tasks and formally define it as an imbalance between policies (IBPs). To address the IBP issue, we propose a dynamic policy balance (DPB) model to balance the learning of each policy by dynamically reweighting the training samples. In addition, current methods for better performance strengthen the exploration of all policies, which leads to disregarding the training differences in the team and reducing learning efficiency. To overcome this drawback, we derive a technique named weighted entropy regularization (WER), a team-level exploration with additional incentives for individuals who exceed the team. DPB and WER are evaluated in homogeneous and heterogeneous tasks, effectively alleviating the imbalanced training problem and improving exploration efficiency. Furthermore, the experimental results show that our models can outperform the state-of-the-art MAPG methods and boast over 12.1% performance gain on average.

In this work, we investigate the potential of large language models (LLMs) based agents to automate data science tasks, with the goal of comprehending task requirements, then building and training the best-fit machine learning models. Despite their widespread success, existing LLM agents are hindered by generating unreasonable experiment plans within this scenario. To this end, we present DS-Agent, a novel automatic framework that harnesses LLM agent and case-based reasoning (CBR). In the development stage, DS-Agent follows the CBR framework to structure an automatic iteration pipeline, which can flexibly capitalize on the expert knowledge from Kaggle, and facilitate consistent performance improvement through the feedback mechanism. Moreover, DS-Agent implements a low-resource deployment stage with a simplified CBR paradigm to adapt past successful solutions from the development stage for direct code generation, significantly reducing the demand on foundational capabilities of LLMs. Empirically, DS-Agent with GPT-4 achieves 100% success rate in the development stage, while attaining 36% improvement on average one pass rate across alternative LLMs in the deployment stage. In both stages, DS-Agent achieves the best rank in performance, costing $1.60 and \$0.13 per run with GPT-4, respectively. Our data and code are open-sourced at https://github.com/guosyjlu/DS-Agent.

In-context learning is a promising approach for offline reinforcement learning (RL) to handle online tasks, which can be achieved by providing task prompts. Recent works demonstrated that in-context RL could emerge with self-improvement in a trial-and-error manner when treating RL tasks as an across-episodic sequential prediction problem. Despite the self-improvement not requiring gradient updates, current works still suffer from high computational costs when the across-episodic sequence increases with task horizons. To this end, we propose an In-context Decision Transformer (IDT) to achieve self-improvement in a high-level trial-and-error manner. Specifically, IDT is inspired by the efficient hierarchical structure of human decision-making and thus reconstructs the sequence to consist of high-level decisions instead of low-level actions that interact with environments. As one high-level decision can guide multi-step low-level actions, IDT naturally avoids excessively long sequences and solves online tasks more efficiently. Experimental results show that IDT achieves state-of-the-art in long-horizon tasks over current in-context RL methods. In particular, the online evaluation time of our IDT is 36$\times$ times faster than baselines in the D4RL benchmark and 27$\times$ times faster in the Grid World benchmark.

Exploring the state space efficiently is a crucial problem in reinforcement learning as it holds significant importance for learning optimal policies. One effective approach involves learning different sub-policies to cover various sub-spaces of the state space, with each sub-policy corresponding to a specific goal. However, the unevenness of the state probability distribution may lead to exploration difficulties in deep reinforcement learning. To overcome this challenge, we propose a Phase State Graph Exploration framework (PSGE), guiding the agent towards more promising directions for exploration. Specifically, we design a graph-based state space exploration framework to separate the combination space into sub-spaces and define the combination space and evaluation criteria for the agent’s sub-policies. In addition, hypernetwork is leveraged to decouple sub-policies and sub-goals, ensuring diversity among the agent’s sub-policies and reward shaping is used to provide dense internal reward signals for policy training, which encourages the agent to learn more efficiently. Experiments on combining control and navigation tasks demonstrate that PSGE performs well in controlling agent across various difficulty level tasks.

The diversity of recommendation is equally crucial as accuracy in improving user experience. Existing studies, e.g., Determinantal Point Process (DPP) and Maximal Marginal Relevance (MMR), employ a greedy paradigm to iteratively select items that optimize both accuracy and diversity. However, prior methods typically exhibit quadratic complexity, limiting their applications to the re-ranking stage and are not applicable to other recommendation stages with a larger pool of candidate items, such as the pre-ranking and ranking stages. In this paper, we propose Contextual Distillation Model (CDM), an efficient recommendation model that addresses diversification, suitable for the deployment in all stages of industrial recommendation pipelines. Specifically, CDM utilizes the candidate items in the same user request as context to enhance the diversification of the results. We propose a contrastive context encoder that employs attention mechanisms to model both positive and negative contexts. For the training of CDM, we compare each target item with its context embedding and utilize the knowledge distillation framework to learn the win probability of each target item under the MMR algorithm, where the teacher is derived from MMR outputs. During inference, ranking is performed through a linear combination of the recommendation and student model scores, ensuring both diversity and efficiency. We perform offline evaluations on two industrial datasets and conduct online A/B test of CDM on the short-video platform KuaiShou. The considerable enhancements observed in both recommendation quality and diversity, as shown by metrics, provide strong superiority for the effectiveness of CDM.

Recent works have shown the remarkable superiority of transformer models in reinforcement learning (RL), where the decision-making problem is formulated as sequential generation. Transformer-based agents could emerge with self-improvement in online environments by providing task contexts, such as multiple trajectories, called in-context RL. However, due to the quadratic computation complexity of attention in transformers, current in-context RL methods suffer from huge computational costs as the task horizon increases. In contrast, the Mamba model is renowned for its efficient ability to process long-term dependencies, which provides an opportunity for in-context RL to solve tasks that require long-term memory. To this end, we first implement Decision Mamba (DM) by replacing the backbone of Decision Transformer (DT). Then, we propose a Decision Mamba-Hybrid (DM-H) with the merits of transformers and Mamba in high-quality prediction and long-term memory. Specifically, DM-H first generates high-value sub-goals from long-term memory through the Mamba model. Then, we use sub-goals to prompt the transformer, establishing high-quality predictions. Experimental results demonstrate that DM-H achieves state-of-the-art in long and short-term tasks, such as D4RL, Grid World, and Tmaze benchmarks. Regarding efficiency, the online testing of DM-H in the long-term task is 28$\times$ times faster than the transformer-based baselines.

Artificial Intelligence (AI) has achieved a wide range of successes in autonomous air combat decision-making recently. Previous research demonstrated that AI-enabled air combat approaches could even acquire beyond human-level capabilities. However, there remains a lack of evidence regarding two major difficulties. First, the existing methods with fixed decision intervals are mostly devoted to solving what to act but merely pay attention to when to act, which occasionally misses optimal decision opportunities. Second, the method of an expert-crafted finite maneuver library leads to a lack of tactics diversity, which is vulnerable to an opponent equipped with new tactics. In view of this, we propose a novel Deep Reinforcement Learning (DRL) and prior knowledge hybrid autonomous air combat tactics discovering algorithm, namely deep E xcitatory-i N hibitory f ACT or I zed maneu VE r ( ENACTIVE ) learning. The algorithm consists of two key modules, i.e., ENHANCE and FACTIVE. Specifically, ENHANCE learns to adjust the air combat decision-making intervals and appropriately seize key opportunities. FACTIVE factorizes maneuvers and then jointly optimizes them with significant tactics diversity increments. Extensive experimental results reveal that the proposed method outperforms state-of-the-art algorithms with a 62% winning rate and further obtains a margin of a 2.85-fold increase in terms of global tactic space coverage. It also demonstrates that a variety of discovered air combat tactics are comparable to human experts’ knowledge.

Traffic congestion is becoming an increasingly prominent problem, and intelligent traffic signal control methods can effectively alleviate it. Recently, there has been a growing trend of applying reinforcement learning to traffic signal control for adaptive signal scheduling. However, most existing methods focus on improving traffic performance while neglecting the issue of scheduling fairness, resulting in long waiting time for some vehicles. Some works attempt to address fairness issues but often sacrifice transport performance. Furthermore, existing methods overlook the challenge of sample efficiency, especially when dealing with diversity-limited traffic data. Therefore, we propose aFairness-aware and sample-Efficient traffic signal control method called FELight. Specifically, we first design a novel fairness metric and integrate it into decision process to penalize cases with high latency by setting a threshold for activating the fairness mechanism. Theoretical comparison with other fairness works proves why and when our fairness could bring advantages. Moreover, counterfactual data augmentation is employed to enrich interaction data, enhancing the sample efficiency of FELight. Self-supervised state representation is introduced to extract informative features from raw states, further improving sample efficiency. Experiments on real traffic datasets demonstrate that FELight provides relatively fairer traffic signal control without compromising performance compared to state-of-the-art approaches. Our codes are available athttps://github.com/dxnbbsw/FELight.

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.

Many graph neural networks (GNNs) are inapplicable when the graph structure representing the node relations is unavailable. Recent studies have shown that this problem can be effectively solved by jointly learning the graph structure and the parameters of GNNs. However, most of these methods learn graphs by using either a Euclidean or hyperbolic metric, which means that the space curvature is assumed to be either constant zero or constant negative. Graph embedding spaces usually have nonconstant curvatures, and thus, such an assumption may produce some obfuscatory nodes, which are improperly embedded and close to multiple categories. In this article, we propose a joint-space graph learning (JSGL) method for GNNs. JSGL learns a graph based on Euclidean embeddings and identifies Euclidean obfuscatory nodes. Then, the graph topology near the identified obfuscatory nodes is refined in hyperbolic space. We also present a theoretical justification of our method for identifying obfuscatory nodes and conduct a series of experiments to test the performance of JSGL. The results show that JSGL outperforms many baseline methods. To obtain more insights, we analyze potential reasons for this superior performance.

Graph neural networks (GNNs) have achieved great success in many fields due to their powerful capabilities of processing graph-structured data. However, most GNNs can only be applied to scenarios where graphs are known, but real-world data are often noisy or even do not have available graph structures. Recently, graph learning has attracted increasing attention in dealing with these problems. In this article, we develop a novel approach to improving the robustness of the GNNs, called composite GNN. Different from existing methods, our method uses composite graphs (C-graphs) to characterize both sample and feature relations. The C-graph is a unified graph that unifies these two kinds of relations, where edges between samples represent sample similarities, and each sample has a tree-based feature graph to model feature importance and combination preference. By jointly learning multiaspect C-graphs and neural network parameters, our method improves the performance of semisupervised node classification and ensures robustness. We conduct a series of experiments to evaluate the performance of our method and the variants of our method that only learn sample relations or feature relations. Extensive experimental results on nine benchmark datasets demonstrate that our proposed method achieves the best performance on almost all the datasets and is robust to feature noises.

In the era of information explosion, named entity recognition (NER) has attracted widespread attention in the field of natural language processing, as it is fundamental to information extraction. Recently, methods of NER based on representation learning, e.g., character embedding and word embedding, have demonstrated promising recognition results. However, existing models only consider partial features derived from words or characters while failing to integrate semantic and syntactic information, e.g., capitalization, inter-word relations, keywords, and lexical phrases, from multilevel perspectives. Intuitively, multilevel features can be helpful when recognizing named entities from complex sentences. In this study, we propose a novel attentive multilevel feature fusion (AMFF) model for NER, which captures the multilevel features in the current context from various perspectives. It consists of four components to, respectively, capture the local character-level (CL), global character-level (CG), local word-level (WL), and global word-level (WG) features in the current context. In addition, we further define document-level features crafted from other sentences to enhance the representation learning of the current context. To this end, we introduce a novel context-aware attentive multilevel feature fusion (CAMFF) model based on AMFF, to fully leverage document-level features from all the previous inputs. The obtained multilevel features are then fused and fed into a bidirectional long short-term memory (BiLSTM)-conditional random field (CRF) network for the final sequence labeling. Extensive experiments on four benchmark datasets demonstrate that our proposed AMFF and CAMFF models outperform a set of state-of-the-art baseline methods and the features learned from multiple levels are complementary.

The popular Proximal Policy Optimization (PPO) algorithm approximates the solution in a clipped policy space. Does there exist better policies outside of this space? By using a novel surrogate objective that employs the sigmoid function (which provides an interesting way of exploration), we found that the answer is "YES", and the better policies are in fact located very far from the clipped space. We show that PPO is insufficient in "off-policyness", according to an off-policy metric called DEON. Our algorithm explores in a much larger policy space than PPO, and it maximizes the Conservative Policy Iteration (CPI) objective better than PPO during training. To the best of our knowledge, all current PPO methods have the clipping operation and optimize in the clipped policy space. Our method is the first of this kind, which advances the understanding of CPI optimization and policy gradient methods. Code is available at https://github.com/raincchio/P3O.

Actor-critic framework has achieved tremendous success in a great many of decision-making scenarios. Nevertheless, when updating the value of new states and actions in the long-term scene, these methods suffer from misestimate problem and gradient variance problem, significantly reducing convergence speed and robustness of the policy. These problems severely limit the application scope of these methods. In this paper, we first proposed QVDDPG, a deep RL algorithm based on the iterative target value update process. The QV learning method alleviates the problem of misestimate by making use of the guidance of Q value and the fast convergence of V value, thus accelerating the convergence speed. In addition, the actor utilizes a constrained balanced gradient and establishes a hidden state for the continuous action space network for the sake of robustness of the model. We give the update relation among the value functions and the constraint conditions of gradient estimation. We measure our method on the PyBullet and achieved state-of-the-art performance. Moreover, we demonstrate that, our method has higher robustness and convergence speed across different tasks compared to other algorithms.

In visual-based Reinforcement Learning (RL), agents often struggle to generalize well to environmental variations in the state space that were not observed during training. The variations can arise in both task-irrelevant features, such as background noise, and task-relevant features, such as robot configurations, that are related to the optimal decisions. To achieve generalization in both situations, agents are required to accurately understand the impact of changed features on the decisions, i.e., establishing the true associations between changed features and decisions in the policy model. However, due to the inherent correlations among features in the state space, the associations between features and decisions become entangled, making it difficult for the policy to distinguish them. To this end, we propose Saliency-Guided Features Decorrelation (SGFD) to eliminate these correlations through sample reweighting. Concretely, SGFD consists of two core techniques: Random Fourier Functions (RFF) and the saliency map. RFF is utilized to estimate the complex non-linear correlations in high-dimensional images, while the saliency map is designed to identify the changed features. Under the guidance of the saliency map, SGFD employs sample reweighting to minimize the estimated correlations related to changed features, thereby achieving decorrelation in visual RL tasks. Our experimental results demonstrate that SGFD can generalize well on a wide range of test environments and significantly outperforms state-of-the-art methods in handling both task-irrelevant variations and task-relevant variations.

Existing fake news detection methods aim to classify a piece of news as true or false and provide veracity explanations, achieving remarkable performances. However, they often tailor automated solutions on manual fact-checked reports, suffering from limited news coverage and debunking delays. When a piece of news has not yet been fact-checked or debunked, certain amounts of relevant raw reports are usually disseminated on various media outlets, containing the wisdom of crowds to verify the news claim and explain its verdict. In this paper, we propose a novel Coarse-to-fine Cascaded Evidence-Distillation (CofCED) neural network for explainable fake news detection based on such raw reports, alleviating the dependency on fact-checked ones. Specifically, we first utilize a hierarchical encoder for web text representation, and then develop two cascaded selectors to select the most explainable sentences for verdicts on top of the selected top-K reports in a coarse-to-fine manner. Besides, we construct two explainable fake news datasets, which is publicly available. Experimental results demonstrate that our model significantly outperforms state-of-the-art detection baselines and generates high-quality explanations from diverse evaluation perspectives.

Multi-agent reinforcement learning has drawn increasing attention in practice, e.g., robotics and automatic driving, as it can explore optimal policies using samples generated by interacting with the environment. However, high reward uncertainty still remains a problem when we want to train a satisfactory model, because obtaining high-quality reward feedback is usually expensive and even infeasible. To handle this issue, previous methods mainly focus on passive reward correction. At the same time, recent active reward estimation methods have proven to be a recipe for reducing the effect of reward uncertainty. In this paper, we propose a novel Distributional Reward Estimation framework for effective Multi-Agent Reinforcement Learning (DRE-MARL). Our main idea is to design the multi-action-branch reward estimation and policy-weighted reward aggregation for stabilized training. Specifically, we design the multi-action-branch reward estimation to model reward distributions on all action branches. Then we utilize reward aggregation to obtain stable updating signals during training. Our intuition is that consideration of all possible consequences of actions could be useful for learning policies. The superiority of the DRE-MARL is demonstrated using benchmark multi-agent scenarios, compared with the SOTA baselines in terms of both effectiveness and robustness.

Predicting the popularity of online videos has many real-world applications, such as recommendation, precise advertising, and edge caching strategies. Despite many efforts have been dedicated to the online video popularity prediction, there still exist several challenges: (1) The meta-data from online videos is usually sparse and noisy, which makes it difficult to learn a stable and robust representation. (2) The influence of content features and temporal features in different life cycles of online videos is dynamically changing, so it is necessary to build a model that can capture the dynamics. (3) Besides, there is a great need to interpret the predictive behavior of the model to assist administrators of video platforms in the subsequent decision-making.

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.

In feeds recommendation, users are able to constantly browse items generated by never-ending feeds using mobile phones. The implicit feedback from users is an important resource for learning to rank, however, building ranking functions from such observed data is recognized to be biased. The presentation of the items will influence the user's judgements and therefore introduces biases. Most previous works in the unbiased learning to rank literature focus on position bias (i.e., an item ranked higher has more chances of being examined and interacted with). By analyzing user behaviors in product feeds recommendation, in this paper, we identify and introduce context bias, which refers to the probability that a user interacting with an item is biased by its surroundings, to unbiased learning to rank. We propose an Unbiased Learning to Rank with Combinational Propensity (ULTR-CP) framework to remove the inherent biases jointly caused by multiple factors. Under this framework, a context-aware position bias model is instantiated to estimate the unified bias considering both position and context biases. In addition to evaluating propensity score estimation approaches by the ranking metrics, we also discuss the evaluation of the propensities directly by checking their balancing properties. Extensive experiments performed on a real e-commerce data set collected from JD.com verify the effectiveness of context bias and illustrate the superiority of ULTR-CP against the state-of-the-art methods.

Stochastic blockmodel (SBM) is a widely used statistical network representation model, with good interpretability, expressiveness, generalization, and flexibility, which has become prevalent and important in the field of network science over the last years. However, learning an optimal SBM for a given network is an NP-hard problem. This results in significant limitations when it comes to applications of SBMs in large-scale networks, because of the significant computational overhead of existing SBM models, as well as their learning methods. Reducing the cost of SBM learning and making it scalable for handling large-scale networks, while maintaining the good theoretical properties of SBM, remains an unresolved problem. In this work, we address this challenging task from a novel perspective of model redefinition. We propose a novel redefined SBM with Poisson distribution and its block-wise learning algorithm that can efficiently analyse large-scale networks. Extensive validation conducted on both artificial and real-world data shows that our proposed method significantly outperforms the state-of-the-art methods in terms of a reasonable trade-off between accuracy and scalability. 1

Many real-world applications reveal difficulties in learning classifiers from imbalanced data. The rising big data era has been witnessing more classification tasks with large-scale but extremely imbalance and low-quality datasets. Most of existing learning methods suffer from poor performance or low computation efficiency under such a scenario. To tackle this problem, we conduct deep investigations into the nature of class imbalance, which reveals that not only the disproportion between classes, but also other difficulties embedded in the nature of data, especially, noises and class overlapping, prevent us from learning effective classifiers. Taking those factors into consideration, we propose a novel framework for imbalance classification that aims to generate a strong ensemble by self-paced harmonizing data hardness via under-sampling. Extensive experiments have shown that this new framework, while being very computationally efficient, can lead to robust performance even under highly overlapping classes and extremely skewed distribution. Note that, our methods can be easily adapted to most of existing learning methods (e.g., C4.5, SVM, GBDT and Neural Network) to boost their performance on imbalanced data.

Named entity recognition (NER) is a fundamental task in the natural language processing (NLP) area. Recently, representation learning methods (e.g., character embedding and word embedding) have achieved promising recognition results. However, existing models only consider partial features derived from words or characters while failing to integrate semantic and syntactic information (e.g., capitalization, inter-word relations, keywords, lexical phrases, etc.) from multi-level perspectives. Intuitively, multi-level features can be helpful when recognizing named entities from complex sentences. In this study, we propose a novel framework called attention-based multi-level feature fusion (AMFF), which is used to capture the multi-level features from different perspectives to improve NER. Our model consists of four components to respectively capture the local character-level, global character-level, local word-level, and global word-level features, which are then fed into a BiLSTM-CRF network for the final sequence labeling. Extensive experimental results on four benchmark datasets show that our proposed model outperforms a set of state-of-the-art baselines.

For quite a long time, effective Beyond-Visual-Range (BVR) air combat tactics can only be discovered by human pilots in the actual combat process. However, due to the lack of actual combat opportunities, making new air combat tactics innovation was generally considered quite difficult. To address this challenge, we first introduced a solely end-to-end Reinforcement Learning (RL) approach for training competitive air combat agents with adversarial self-play from scratch in a high fidelity air combat simulation environment during training. Furthermore, a Key Air Combat Event Reward Shaping (KAERS) mechanism was proposed to provide sparse but objective shaped rewards beyond episodic win/lose signal to accelerate the initial machine learning process. Experimental results showed that multiple valuable air combat tactical behaviors emerged progressively. We hope this study could be extended to the future of air combat machine intelligence research.

During an epidemic, the spatial, temporal and demographic patterns of disease transmission are determined by multiple factors. In addition to the physiological properties of the pathogens and hosts, the social contact of the host population, which characterizes the reciprocal exposures of individuals to infection according to their demographic structure and various social activities, are also pivotal to understanding and predicting the prevalence of infectious diseases. How social contact is measured will affect the extent to which we can forecast the dynamics of infections in the real world. Most current work focuses on modeling the spatial patterns of static social contact. In this work, we use a novel perspective to address the problem of how to characterize and measure dynamic social contact during an epidemic. We propose an epidemic-model-based tensor deconvolution framework in which the spatiotemporal patterns of social contact are represented by the factors of the tensors. These factors can be discovered using a tensor deconvolution procedure with the integration of epidemic models based on rich types of data, mainly heterogeneous outbreak surveillance data, socio-demographic census data and physiological data from medical reports. Using reproduction models that include SIR/SIS/SEIR/SEIS models as case studies, the efficacy and applications of the proposed framework are theoretically analyzed, empirically validated and demonstrated through a set of rigorous experiments using both synthetic and real-world data.

During an epidemic, the spatial, temporal and demographical patterns of disease transmission are determined by multiple factors. Besides the physiological properties of pathogenes and hosts, the social contacts of host population, which characterize individuals' reciprocal exposures of infection in view of demographical structures and various social activities, are also pivotal to understand and further predict the prevalence of infectious diseases. The means of measuring social contacts will dominate the extent how precisely we can forecast the dynamics of infections in the real world. Most current works focus their efforts on modeling the spatial patterns of static social contacts. In this work, we address the problem on how to characterize and measure dynamical social contacts during an epidemic from a novel perspective. We propose an epidemic-model-based tensor deconvolution framework to address this issue, in which the spatiotemporal patterns of social contacts are represented by the factors of tensors, which can be discovered by a tensor deconvolution procedure with an integration of epidemic models from rich types of data, mainly including heterogeneous outbreak surveillance, social-demographic census and physiological data from medical reports. Taking SIR model as a case study, the efficacy of the proposed method is theoretically analyzed and empirically validated through a set of rigorous experiments on both synthetic and real-world data.