Hao Hu

Kimi Team, Tongtong Bai, Yifan Bai et al.

We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel agent orchestration framework that dynamically decomposes complex tasks into heterogeneous sub-problems and executes them concurrently. Extensive evaluations show that Kimi K2.5 achieves state-of-the-art results across various domains including coding, vision, reasoning, and agentic tasks. Agent Swarm also reduces latency by up to $4.5\times$ over single-agent baselines. We release the post-trained Kimi K2.5 model checkpoint to facilitate future research and real-world applications of agentic intelligence.

Bowen Wang, Dunjie Lu, Junli Wang et al.

Reinforcement learning with verifiable rewards (RLVR) has driven breakthroughs in domains such as math, tool-use, and software engineering, yet its extension to computer-use agents (CUAs) has been bottlenecked by the scarcity of scalable training data with deterministic rewards. Constructing such data for CUAs requires consistent task instruction, executable environment, and verifiable reward. However, hand-curated benchmarks achieve high reward fidelity but cover few applications and LLM-as-judge-based datasets scale broadly but lack reliable verification. We present CUA-Gym, a scalable pipeline that co-generates task instructions, environment states, and reward functions. Concretely, a Generator agent constructs the initial and golden environment states, and a separate Discriminator agent writes the reward function from the task specification. An orchestrator agent drives the two through iterative rounds upon execution. Generated tuples then pass a final filter combining LLM majority voting and agent rollouts, ensuring quality beyond the per-task adversarial loop. To address the scarcity of training environments, we further synthesize CUA-Gym-Hub, a broad suite of high-fidelity mock web applications grounded in real-world software-use distributions, expanding the scale of CUA RLVR data by magnitude. Using this pipeline, we construct CUA-Gym, a dataset of 32,112 verified RLVR training tuples grounded in 110 environments. Trained with GSPO on CUA-Gym, our CUA-Gym-A3B and CUA-Gym-A17B achieve 62.1% and 72.6% on OSWorld-Verified, outperforming prior open-source CUAs at comparable scales, with performance scaling smoothly in both data volume and environment diversity. The same checkpoints also improve on the held-out WebArena benchmark, indicating transfer beyond the training environments. We will open-source the full synthesis pipeline, dataset, CUA-Gym-Hub environments, and models.

Jae Yong Lee, Daniel Scharstein, Akash Bapat et al.

We present Ego-1K, a large-scale collection of time-synchronized egocentric multiview videos designed to advance neural 3D video synthesis and dynamic scene understanding. The dataset contains nearly 1,000 short egocentric videos captured with a custom rig with 12 synchronized cameras surrounding a 4-camera VR headset worn by the user. Scene content focuses on hand motions and hand-object interactions in different settings. We describe rig design, data processing, and calibration. Our dataset enables new ways to benchmark egocentric scene reconstruction methods, an important research area as smart glasses with multiple cameras become omnipresent. Our experiments demonstrate that our dataset presents unique challenges for existing 3D and 4D novel view synthesis methods due to large disparities and image motion caused by close dynamic objects and rig egomotion. Our dataset supports future research in this challenging domain. It is available at https://huggingface.co/datasets/facebook/ego-1k.

Zhihan Liu, Hao Hu, Shenao Zhang et al.

Large language models (LLMs) demonstrate impressive reasoning abilities, but translating reasoning into actions in the real world remains challenging. In particular, it remains unclear how to complete a given task provably within a minimum number of interactions with the external environment, e.g., through an internal mechanism of reasoning. To this end, we propose a principled framework with provable regret guarantees to orchestrate reasoning and acting, which we call "reason for future, act for now" (\texttt{RAFA}). Specifically, we design a prompt template for reasoning that learns from the memory buffer and plans a future trajectory over a long horizon ("reason for future"). At each step, the LLM agent takes the initial action of the planned trajectory ("act for now"), stores the collected feedback in the memory buffer, and reinvokes the reasoning routine to replan the future trajectory from the new state. The key idea is to cast reasoning in LLMs as learning and planning in Bayesian adaptive Markov decision processes (MDPs). Correspondingly, we prompt LLMs to form an updated posterior of the unknown environment from the memory buffer (learning) and generate an optimal trajectory for multiple future steps that maximizes a value function (planning). The learning and planning subroutines are performed in an "in-context" manner to emulate the actor-critic update for MDPs. Our theoretical analysis proves that the novel combination of long-term reasoning and short-term acting achieves a $\sqrt{T}$ regret. Here, $T$ denotes the number of online interactions. In particular, the regret bound highlights an intriguing interplay between the prior knowledge obtained through pretraining and the uncertainty reduction achieved by reasoning and acting. Our empirical validation shows that it outperforms various existing frameworks and achieves nearly perfect scores on a few benchmarks.

Hao Hu, Yiqin Yang, Qianchuan Zhao et al.

Offline reinforcement learning (RL) enables effective learning from previously collected data without exploration, which shows great promise in real-world applications when exploration is expensive or even infeasible. The discount factor, $γ$, plays a vital role in improving online RL sample efficiency and estimation accuracy, but the role of the discount factor in offline RL is not well explored. This paper examines two distinct effects of $γ$ in offline RL with theoretical analysis, namely the regularization effect and the pessimism effect. On the one hand, $γ$ is a regulator to trade-off optimality with sample efficiency upon existing offline techniques. On the other hand, lower guidance $γ$ can also be seen as a way of pessimism where we optimize the policy's performance in the worst possible models. We empirically verify the above theoretical observation with tabular MDPs and standard D4RL tasks. The results show that the discount factor plays an essential role in the performance of offline RL algorithms, both under small data regimes upon existing offline methods and in large data regimes without other conservative methods.

Yiqin Yang, Hao Hu, Wenzhe Li et al.

Offline reinforcement learning (RL) enables the agent to effectively learn from logged data, which significantly extends the applicability of RL algorithms in real-world scenarios where exploration can be expensive or unsafe. Previous works have shown that extracting primitive skills from the recurring and temporally extended structures in the logged data yields better learning. However, these methods suffer greatly when the primitives have limited representation ability to recover the original policy space, especially in offline settings. In this paper, we give a quantitative characterization of the performance of offline hierarchical learning and highlight the importance of learning lossless primitives. To this end, we propose to use a \emph{flow}-based structure as the representation for low-level policies. This allows us to represent the behaviors in the dataset faithfully while keeping the expression ability to recover the whole policy space. We show that such lossless primitives can drastically improve the performance of hierarchical policies. The experimental results and extensive ablation studies on the standard D4RL benchmark show that our method has a good representation ability for policies and achieves superior performance in most tasks.

Hao Hu, Yiqin Yang, Qianchuan Zhao et al.

Self-supervised methods have become crucial for advancing deep learning by leveraging data itself to reduce the need for expensive annotations. However, the question of how to conduct self-supervised offline reinforcement learning (RL) in a principled way remains unclear. In this paper, we address this issue by investigating the theoretical benefits of utilizing reward-free data in linear Markov Decision Processes (MDPs) within a semi-supervised setting. Further, we propose a novel, Provable Data Sharing algorithm (PDS) to utilize such reward-free data for offline RL. PDS uses additional penalties on the reward function learned from labeled data to prevent overestimation, ensuring a conservative algorithm. Our results on various offline RL tasks demonstrate that PDS significantly improves the performance of offline RL algorithms with reward-free data. Overall, our work provides a promising approach to leveraging the benefits of unlabeled data in offline RL while maintaining theoretical guarantees. We believe our findings will contribute to developing more robust self-supervised RL methods.

Hao Hu, Marie-José Huguet, Mohamed Siala

The growing interest in explainable artificial intelligence (XAI) for critical decision making motivates the need for interpretable machine learning (ML) models. In fact, due to their structure (especially with small sizes), these models are inherently understandable by humans. Recently, several exact methods for computing such models are proposed to overcome weaknesses of traditional heuristic methods by providing more compact models or better prediction quality. Despite their compressed representation of Boolean functions, Binary decision diagrams (BDDs) did not gain enough interest as other interpretable ML models. In this paper, we first propose SAT-based models for learning optimal BDDs (in terms of the number of features) that classify all input examples. Then, we lift the encoding to a MaxSAT model to learn optimal BDDs in limited depths, that maximize the number of examples correctly classified. Finally, we tackle the fragmentation problem by introducing a method to merge compatible subtrees for the BDDs found via the MaxSAT model. Our empirical study shows clear benefits of the proposed approach in terms of prediction quality and intrepretability (i.e., lighter size) compared to the state-of-the-art approaches.

Yiqi Sun, Zhengxin Shi, Jianshen Zhang et al. · ibm-research

In time series forecasting, decomposition-based algorithms break aggregate data into meaningful components and are therefore appreciated for their particular advantages in interpretability. Recent algorithms often combine machine learning (hereafter ML) methodology with decomposition to improve prediction accuracy. However, incorporating ML is generally considered to sacrifice interpretability inevitably. In addition, existing hybrid algorithms usually rely on theoretical models with statistical assumptions and focus only on the accuracy of aggregate predictions, and thus suffer from accuracy problems, especially in component estimates. In response to the above issues, this research explores the possibility of improving accuracy without losing interpretability in time series forecasting. We first quantitatively define interpretability for data-driven forecasts and systematically review the existing forecasting algorithms from the perspective of interpretability. Accordingly, we propose the W-R algorithm, a hybrid algorithm that combines decomposition and ML from a novel perspective. Specifically, the W-R algorithm replaces the standard additive combination function with a weighted variant and uses ML to modify the estimates of all components simultaneously. We mathematically analyze the theoretical basis of the algorithm and validate its performance through extensive numerical experiments. In general, the W-R algorithm outperforms all decomposition-based and ML benchmarks. Based on P50_QL, the algorithm relatively improves by 8.76% in accuracy on the practical sales forecasts of JD.com and 77.99% on a public dataset of electricity loads. This research offers an innovative perspective to combine the statistical and ML algorithms, and JD.com has implemented the W-R algorithm to make accurate sales predictions and guide its marketing activities.

Kimi Team, Angang Du, Bohong Yin et al. · pku, tsinghua

We present Kimi-VL, an efficient open-source Mixture-of-Experts (MoE) vision-language model (VLM) that offers advanced multimodal reasoning, long-context understanding, and strong agent capabilities - all while activating only 2.8B parameters in its language decoder (Kimi-VL-A3B). Kimi-VL demonstrates strong performance across challenging domains: as a general-purpose VLM, Kimi-VL excels in multi-turn agent tasks (e.g., OSWorld), matching flagship models. Furthermore, it exhibits remarkable capabilities across diverse challenging vision language tasks, including college-level image and video comprehension, OCR, mathematical reasoning, and multi-image understanding. In comparative evaluations, it effectively competes with cutting-edge efficient VLMs such as GPT-4o-mini, Qwen2.5-VL-7B, and Gemma-3-12B-IT, while surpassing GPT-4o in several key domains. Kimi-VL also advances in processing long contexts and perceiving clearly. With a 128K extended context window, Kimi-VL can process diverse long inputs, achieving impressive scores of 64.5 on LongVideoBench and 35.1 on MMLongBench-Doc. Its native-resolution vision encoder, MoonViT, further allows it to see and understand ultra-high-resolution visual inputs, achieving 83.2 on InfoVQA and 34.5 on ScreenSpot-Pro, while maintaining lower computational cost for common tasks. Building upon Kimi-VL, we introduce an advanced long-thinking variant: Kimi-VL-Thinking-2506. Developed through long chain-of-thought (CoT) supervised fine-tuning (SFT) and reinforcement learning (RL), the latest model exhibits strong long-horizon reasoning capabilities (64.0 on MMMU, 46.3 on MMMU-Pro, 56.9 on MathVision, 80.1 on MathVista, 65.2 on VideoMMMU) while obtaining robust general abilities. Code and models are publicly accessible at https://github.com/MoonshotAI/Kimi-VL.

Hao Hu, Yiqin Yang, Jianing Ye et al.

Reward-free data is abundant and contains rich prior knowledge of human behaviors, but it is not well exploited by offline reinforcement learning (RL) algorithms. In this paper, we propose UBER, an unsupervised approach to extract useful behaviors from offline reward-free datasets via diversified rewards. UBER assigns different pseudo-rewards sampled from a given prior distribution to different agents to extract a diverse set of behaviors, and reuse them as candidate policies to facilitate the learning of new tasks. Perhaps surprisingly, we show that rewards generated from random neural networks are sufficient to extract diverse and useful behaviors, some even close to expert ones. We provide both empirical and theoretical evidence to justify the use of random priors for the reward function. Experiments on multiple benchmarks showcase UBER's ability to learn effective and diverse behavior sets that enhance sample efficiency for online RL, outperforming existing baselines. By reducing reliance on human supervision, UBER broadens the applicability of RL to real-world scenarios with abundant reward-free data.

Kimi Team, Yifan Bai, Yiping Bao et al. · tsinghua

We introduce Kimi K2, a Mixture-of-Experts (MoE) large language model with 32 billion activated parameters and 1 trillion total parameters. We propose the MuonClip optimizer, which improves upon Muon with a novel QK-clip technique to address training instability while enjoying the advanced token efficiency of Muon. Based on MuonClip, K2 was pre-trained on 15.5 trillion tokens with zero loss spike. During post-training, K2 undergoes a multi-stage post-training process, highlighted by a large-scale agentic data synthesis pipeline and a joint reinforcement learning (RL) stage, where the model improves its capabilities through interactions with real and synthetic environments. Kimi K2 achieves state-of-the-art performance among open-source non-thinking models, with strengths in agentic capabilities. Notably, K2 obtains 66.1 on Tau2-Bench, 76.5 on ACEBench (En), 65.8 on SWE-Bench Verified, and 47.3 on SWE-Bench Multilingual -- surpassing most open and closed-sourced baselines in non-thinking settings. It also exhibits strong capabilities in coding, mathematics, and reasoning tasks, with a score of 53.7 on LiveCodeBench v6, 49.5 on AIME 2025, 75.1 on GPQA-Diamond, and 27.1 on OJBench, all without extended thinking. These results position Kimi K2 as one of the most capable open-source large language models to date, particularly in software engineering and agentic tasks. We release our base and post-trained model checkpoints to facilitate future research and applications of agentic intelligence.

Xinyuan Wang, Bowen Wang, Dunjie Lu et al. · cmu

Vision-language models have demonstrated impressive capabilities as computer-use agents (CUAs) capable of automating diverse computer tasks. As their commercial potential grows, critical details of the most capable CUA systems remain closed. As these agents will increasingly mediate digital interactions and execute consequential decisions on our behalf, the research community needs access to open CUA frameworks to study their capabilities, limitations, and risks. To bridge this gap, we propose OpenCUA, a comprehensive open-source framework for scaling CUA data and foundation models. Our framework consists of: (1) an annotation infrastructure that seamlessly captures human computer-use demonstrations; (2) AgentNet, the first large-scale computer-use task dataset spanning 3 operating systems and 200+ applications and websites; (3) a scalable pipeline that transforms demonstrations into state-action pairs with reflective long Chain-of-Thought reasoning that sustain robust performance gains as data scales. Our end-to-end agent models demonstrate strong performance across CUA benchmarks. In particular, OpenCUA-72B achieves an average success rate of 45.0% on OSWorld-Verified, establishing a new state-of-the-art (SOTA) among open-source models. Further analysis confirms that our approach generalizes well across domains and benefits significantly from increased test-time computation. We release our annotation tool, datasets, code, and models to build open foundations for further CUA research.

Kimi Team, Angang Du, Bofei Gao et al. · pku, tsinghua

Language model pretraining with next token prediction has proved effective for scaling compute but is limited to the amount of available training data. Scaling reinforcement learning (RL) unlocks a new axis for the continued improvement of artificial intelligence, with the promise that large language models (LLMs) can scale their training data by learning to explore with rewards. However, prior published work has not produced competitive results. In light of this, we report on the training practice of Kimi k1.5, our latest multi-modal LLM trained with RL, including its RL training techniques, multi-modal data recipes, and infrastructure optimization. Long context scaling and improved policy optimization methods are key ingredients of our approach, which establishes a simplistic, effective RL framework without relying on more complex techniques such as Monte Carlo tree search, value functions, and process reward models. Notably, our system achieves state-of-the-art reasoning performance across multiple benchmarks and modalities -- e.g., 77.5 on AIME, 96.2 on MATH 500, 94-th percentile on Codeforces, 74.9 on MathVista -- matching OpenAI's o1. Moreover, we present effective long2short methods that use long-CoT techniques to improve short-CoT models, yielding state-of-the-art short-CoT reasoning results -- e.g., 60.8 on AIME, 94.6 on MATH500, 47.3 on LiveCodeBench -- outperforming existing short-CoT models such as GPT-4o and Claude Sonnet 3.5 by a large margin (up to +550%).

Wen Wang, Xiangchen Wu, Liang Wang et al.

The Capacitated Vehicle Routing Problem (CVRP) is a fundamental NP-hard problem with broad applications in logistics and transportation. Real-world CVRPs often involve diverse objectives and complex constraints, such as time windows or backhaul requirements, motivating the development of a unified solution framework. Recent reinforcement learning (RL) approaches have shown promise in combinatorial optimization, yet they rely on end-to-end learning and lack explicit problem-solving knowledge, limiting solution quality. In this paper, we propose a knowledge-embedded framework inspired by the Route-First Cluster-Second heuristics. It incorporates knowledge at two levels: (1) decomposing CVRPs into the route-first and cluster-second subproblems, and (2) leveraging dynamic programming to solve the second subproblem, whose results guide the RL-based constructive solver to solve the first problem. To mitigate partial observability caused by problem decomposition, we introduce a unified history-enhanced context processing module. Extensive experiments show that this framework achieves superior solution quality compared with state-of-the-art learning-based methods, with a smaller gap to classical heuristics, demonstrating strong generalization across diverse CVRP variants.

Yixiao Peng, Hao Hu, Feiyang Li et al.

While virtualization and resource pooling empower cloud networks with structural flexibility and elastic scalability, they inevitably expand the attack surface and challenge cyber resilience. Reinforcement Learning (RL)-based defense strategies have been developed to optimize resource deployment and isolation policies under adversarial conditions, aiming to enhance system resilience by maintaining and restoring network availability. However, existing approaches lack robustness as they require retraining to adapt to dynamic changes in network structure, node scale, attack strategies, and attack intensity. Furthermore, the lack of Human-in-the-Loop (HITL) support limits interpretability and flexibility. To address these limitations, we propose CyberOps-Bots, a hierarchical multi-agent reinforcement learning framework empowered by Large Language Models (LLMs). Inspired by MITRE ATT&CK's Tactics-Techniques model, CyberOps-Bots features a two-layer architecture: (1) An upper-level LLM agent with four modules--ReAct planning, IPDRR-based perception, long-short term memory, and action/tool integration--performs global awareness, human intent recognition, and tactical planning; (2) Lower-level RL agents, developed via heterogeneous separated pre-training, execute atomic defense actions within localized network regions. This synergy preserves LLM adaptability and interpretability while ensuring reliable RL execution. Experiments on real cloud datasets show that, compared to state-of-the-art algorithms, CyberOps-Bots maintains network availability 68.5% higher and achieves a 34.7% jumpstart performance gain when shifting the scenarios without retraining. To our knowledge, this is the first study to establish a robust LLM-RL framework with HITL support for cloud defense. We will release our framework to the community, facilitating the advancement of robust and autonomous defense in cloud networks.

Hao Hu, Haijun Yu

Hermite polynomials and functions have extensive applications in scientific and engineering problems. Although it is recognized that employing the scaled Hermite functions rather than the standard ones can remarkably enhance the approximation performance, the understanding of the scaling factor remains insufficient. Due to the lack of theoretical analysis, recent publications still cast doubt on whether the Hermite spectral method is inferior to other methods. To dispel this doubt, we show in this article that the inefficiency of the Hermite spectral method comes from the imbalance in the decay speed of the objective function within the spatial and frequency domains. Proper scaling can render the Hermite spectral methods comparable to other methods. To make it solid, we propose a novel error analysis framework for the scaled Hermite approximation. Taking the $L^2$ projection error as an example, our framework illustrates that there are three different components of errors: the spatial truncation error, the frequency truncation error, and the Hermite spectral approximation error. Through this perspective, finding the optimal scaling factor is equivalent to balancing the spatial and frequency truncation errors. As applications, we show that geometric convergence can be recovered by proper scaling for a class of functions. Furthermore, we show that proper scaling can double the convergence order for smooth functions with algebraic decay. The perplexing pre-asymptotic sub-geometric convergence when approximating algebraic decay functions can be perfectly explained by this framework.

Sicong Pan, Hao Hu, Xuying Huang et al.

Object-centric view planning is a core component of active geometric 3D reconstruction in robotics, yet existing evaluations often conflate object complexity, planning difficulty, budget assumptions, and physical reachability constraints. As a result, conclusions drawn from idealized view-planning evaluations may not reliably predict performance under realistic reconstruction settings. We introduce ObjView-Bench, an evaluation framework for rethinking difficulty and deployment in object-centric view planning. First, we disentangle three quantities underlying view-planning evaluation: omnidirectional self-occlusion as an object-side attribute, observation saturation difficulty, and protocol-dependent planning difficulty defined through a set-cover formulation. This separation supports controlled dataset construction, analysis of slow-saturation objects, and a case study showing that planning difficulty-aware sampling can improve learned view planners. Second, we design deployment-oriented evaluation protocols that reveal how budget regimes and reachable-view constraints alter method behavior. Across classical, learned, and hybrid planners, ObjView-Bench shows that difficulty, budget, and reachability constraints substantially change method rankings and failure modes.

Zhihan Chen, Peng Lin, Hao Hu et al.

As a broadly applied technique in numerous optimization problems, recently, local search has been employed to solve Pseudo-Boolean Optimization (PBO) problem. A representative local search solver for PBO is LSPBO. In this paper, firstly, we improve LSPBO by a dynamic scoring mechanism, which dynamically strikes a balance between score on hard constraints and score on the objective function. Moreover, on top of this improved LSPBO , we develop the first parallel local search PBO solver. The main idea is to share good solutions among different threads to guide the search, by maintaining a pool of feasible solutions. For evaluating solutions when updating the pool, we propose a function that considers both the solution quality and the diversity of the pool. Furthermore, we calculate the polarity density in the pool to enhance the scoring function of local search. Our empirical experiments show clear benefits of the proposed parallel approach, making it competitive with the parallel version of the famous commercial solver Gurobi.

Xia Yang, Xuanyi Zhang, Hao Hu et al.

Large language models now achieve high final-answer accuracy on mathematical reasoning benchmarks, but accuracy alone does not capture reasoning flexibility. We introduce a strategy-level evaluation framework instantiated on 80 AMC 10/12 and AIME problems with 217 AoPS-derived reference strategy families. Model outputs are annotated for strategy identity, validity, and correctness using dual-AI coding with human adjudication. Across four frontier models, we find a pronounced decoupling between answer accuracy and strategy diversity. Under a single-solution prompt, all models achieve high accuracy (95%-100%), but under a multiple-strategy prompt they recover substantially fewer strategies than the human reference set. Gemini, DeepSeek, GPT, and Claude generate 184, 152, 151, and 110 distinct valid strategies, respectively, with the largest gaps in Geometry and Number Theory. The models collectively produce 50 benchmark-novel valid strategies, indicating both incomplete coverage of human strategies and some capacity for alternative reasoning. A repeated-run robustness check on 20 problems shows diminishing gains in discovered strategies, with the strongest model recovering only 39 of 55 AoPS-reference strategies (71%) after three runs. These findings position strategy diversity as a complementary dimension for evaluating mathematical reasoning beyond answer correctness.

Lingjie Zhao, Xue Yu, Yongzhi Qi et al.

As the pursuit of synergy between Artificial Intelligence (AI) and Operations Research (OR) gains momentum in handling complex inventory systems, a critical challenge persists: how to effectively reconcile AI's adaptive perception with OR's structural rigor. To bridge this gap, we propose a novel OR-Guided "Pretrain-then-Reinforce" framework. To provide structured guidance, we propose a simulation-augmented OR model that generates high-quality reference decisions, implicitly capturing complex business constraints and managerial preferences. Leveraging these OR-derived decisions as foundational training labels, we design a domain-informed deep learning foundation model to establish foundational decision-making capabilities, followed by a reinforcement learning (RL) fine-tuning stage. Uniquely, we position RL as a deep alignment mechanism that enables the AI agent to internalize the optimality principles of OR, while simultaneously leveraging exploration for general policy refinement and allowing expert guidance for scenario-specific adaptation (e.g., promotional events). Validated through extensive numerical experiments and a field deployment at JD.com augmented by a Difference-in-Differences (DiD) analysis, our model significantly outperforms incumbent industrial practices, delivering real-world gains of a 5.27-day reduction in turnover and a 2.29% increase in in-stock rates, alongside a 29.95% decrease in holding costs. Contrary to the prevailing trend of brute-force model scaling, our study demonstrates that a lightweight, domain-informed model can deliver state-of-the-art performance and robust transferability when guided by structured OR logic. This approach offers a scalable and cost-effective paradigm for intelligent supply chain management, highlighting the value of deeply aligning AI with OR.

Yiqin Yang, Hao Hu, Yihuan Mao et al.

Preference-based reinforcement learning (PbRL) can help avoid sophisticated reward designs and align better with human intentions, showing great promise in various real-world applications. However, obtaining human feedback for preferences can be expensive and time-consuming, which forms a strong barrier for PbRL. In this work, we address the problem of low query efficiency in offline PbRL, pinpointing two primary reasons: inefficient exploration and overoptimization of learned reward functions. In response to these challenges, we propose a novel algorithm, \textbf{O}ffline \textbf{P}b\textbf{R}L via \textbf{I}n-\textbf{D}ataset \textbf{E}xploration (OPRIDE), designed to enhance the query efficiency of offline PbRL. OPRIDE consists of two key features: a principled exploration strategy that maximizes the informativeness of the queries and a discount scheduling mechanism aimed at mitigating overoptimization of the learned reward functions. Through empirical evaluations, we demonstrate that OPRIDE significantly outperforms prior methods, achieving strong performance with notably fewer queries. Moreover, we provide theoretical guarantees of the algorithm's efficiency. Experimental results across various locomotion, manipulation, and navigation tasks underscore the efficacy and versatility of our approach.

Runpeng Xie, Quanwei Wang, Hao Hu et al.

Comprehending natural language and following human instructions are critical capabilities for intelligent agents. However, the flexibility of linguistic instructions induces substantial ambiguity across language-conditioned tasks, severely degrading algorithmic performance. To address these limitations, we present a novel method named DAIL (Distributional Aligned Learning), featuring two key components: distributional policy and semantic alignment. Specifically, we provide theoretical results that the value distribution estimation mechanism enhances task differentiability. Meanwhile, the semantic alignment module captures the correspondence between trajectories and linguistic instructions. Extensive experimental results on both structured and visual observation benchmarks demonstrate that DAIL effectively resolves instruction ambiguities, achieving superior performance to baseline methods. Our implementation is available at https://github.com/RunpengXie/Distributional-Aligned-Learning.

Pengbo Shen, Yaqing Wang, Ni Mu et al.

Evaluating large language models (LLMs) in complex decision-making is essential for advancing AI's ability for strategic planning and real-time adaptation. However, existing benchmarks for tasks like StarCraft II fail to capture the game's full complexity, such as its complete game context, diverse action spaces, and all playable races. To address this gap, we present SC2Arena, a benchmark that fully supports all playable races, low-level action spaces, and optimizes text-based observations to tackle spatial reasoning challenges. Complementing this, we introduce StarEvolve, a hierarchical framework that integrates strategic planning with tactical execution, featuring iterative self-correction and continuous improvement via fine-tuning on high-quality gameplay data. Its key components include a Planner-Executor-Verifier structure to break down gameplay, and a scoring system for selecting high-quality training samples. Comprehensive analysis using SC2Arena provides valuable insights into developing generalist agents that were not possible with previous benchmarks. Experimental results also demonstrate that our proposed StarEvolve achieves superior performance in strategic planning. Our code, environment, and algorithms are publicly available.

Bo Liu, Liqiang Yu, Chang Che et al.

This paper focuses on the analysis of the application effectiveness of the integration of deep learning and computer vision technologies. Deep learning achieves a historic breakthrough by constructing hierarchical neural networks, enabling end-to-end feature learning and semantic understanding of images. The successful experiences in the field of computer vision provide strong support for training deep learning algorithms. The tight integration of these two fields has given rise to a new generation of advanced computer vision systems, significantly surpassing traditional methods in tasks such as machine vision image classification and object detection. In this paper, typical image classification cases are combined to analyze the superior performance of deep neural network models while also pointing out their limitations in generalization and interpretability, proposing directions for future improvements. Overall, the efficient integration and development trend of deep learning with massive visual data will continue to drive technological breakthroughs and application expansion in the field of computer vision, making it possible to build truly intelligent machine vision systems. This deepening fusion paradigm will powerfully promote unprecedented tasks and functions in computer vision, providing stronger development momentum for related disciplines and industries.

Hao Hu, Shaowei Cai

The growing interest in Explainable Artificial Intelligence (XAI) motivates promising studies of computing optimal Interpretable Machine Learning models, especially decision trees. Such models generally provide optimality in compact size or empirical accuracy. Recent works focus on improving efficiency due to the natural scalability issue. The application of such models to practical problems is quite limited. As an emerging problem in circuit design, Approximate Logic Synthesis (ALS) aims to reduce circuit complexity by sacrificing correctness. Recently, multiple heuristic machine learning methods have been applied in ALS, which learns approximated circuits from samples of input-output pairs. In this paper, we propose a new ALS methodology realizing the approximation via learning optimal decision trees in empirical accuracy. Compared to previous heuristic ALS methods, the guarantee of optimality achieves a more controllable trade-off between circuit complexity and accuracy. Experimental results show clear improvements in our methodology in the quality of approximated designs (circuit complexity and accuracy) compared to the state-of-the-art approaches.

Zhaolong Ning, Hao Hu, Xiaojie Wang et al.

Intelligent reflecting surface (IRS)-assisted unmanned aerial vehicle (UAV) communications are expected to alleviate the load of ground base stations in a cost-effective way. Existing studies mainly focus on the deployment and resource allocation of a single IRS instead of multiple IRSs, whereas it is extremely challenging for joint multi-IRS multi-user association in UAV communications with constrained reflecting resources and dynamic scenarios. To address the aforementioned challenges, we propose a new optimization algorithm for joint IRS-user association, trajectory optimization of UAVs, successive interference cancellation (SIC) decoding order scheduling and power allocation to maximize system energy efficiency. We first propose an inverse soft-Q learning-based algorithm to optimize multi-IRS multi-user association. Then, SCA and Dinkelbach-based algorithm are leveraged to optimize UAV trajectory followed by the optimization of SIC decoding order scheduling and power allocation. Finally, theoretical analysis and performance results show significant advantages of the designed algorithm in convergence rate and energy efficiency.

Yanming Mu, Hao Hu, Feiyang Li et al.

Retrieval-Augmented Generation (RAG) significantly mitigates the hallucinations and domain knowledge deficiency in large language models by incorporating external knowledge bases. However, the multi-module architecture of RAG introduces complex system-level security vulnerabilities. Guided by the RAG workflow, this paper analyzes the underlying vulnerability mechanisms and systematically categorizes core threat vectors such as data poisoning, adversarial attacks, and membership inference attacks. Based on this threat assessment, we construct a taxonomy of RAG defense technologies from a dual perspective encompassing both input and output stages. The input-side analysis reviews data protection mechanisms including dynamic access control, homomorphic encryption retrieval, and adversarial pre-filtering. The output-side examination summarizes advanced leakage prevention techniques such as federated learning isolation, differential privacy perturbation, and lightweight data sanitization. To establish a unified benchmark for future experimental design, we consolidate authoritative test datasets, security standards, and evaluation frameworks. To the best of our knowledge, this paper presents the first end-to-end survey dedicated to the security of RAG systems. Distinct from existing literature that isolates specific vulnerabilities, we systematically map the entire pipeline-providing a unified analysis of threat models, defense mechanisms, and evaluation benchmarks. By enabling deep insights into potential risks, this work seeks to foster the development of highly robust and trustworthy next-generation RAG systems.

Yiqin Yang, Xu Yang, Yuhua Jiang et al.

In the realm of multi-agent systems, the challenge of \emph{partial observability} is a critical barrier to effective coordination and decision-making. Existing approaches, such as belief state estimation and inter-agent communication, often fall short. Belief-based methods are limited by their focus on past experiences without fully leveraging global information, while communication methods often lack a robust model to effectively utilize the auxiliary information they provide. To solve this issue, we propose Global State Diffusion Algorithm~(GlobeDiff) to infer the global state based on the local observations. By formulating the state inference process as a multi-modal diffusion process, GlobeDiff overcomes ambiguities in state estimation while simultaneously inferring the global state with high fidelity. We prove that the estimation error of GlobeDiff under both unimodal and multi-modal distributions can be bounded. Extensive experimental results demonstrate that GlobeDiff achieves superior performance and is capable of accurately inferring the global state.

Yongzhi Qi, Hao Hu, Dazhou Lei et al.

Time series neural networks perform exceptionally well in real-world applications but encounter challenges such as limited scalability, poor generalization, and suboptimal zero-shot performance. Inspired by large language models, there is interest in developing large time series models (LTM) to address these issues. However, current methods struggle with training complexity, adapting human feedback, and achieving high predictive accuracy. We introduce TimeHF, a novel pipeline for creating LTMs with 6 billion parameters, incorporating human feedback. We use patch convolutional embedding to capture long time series information and design a human feedback mechanism called time-series policy optimization. Deployed in JD.com's supply chain, TimeHF handles automated replenishment for over 20,000 products, improving prediction accuracy by 33.21% over existing methods. This work advances LTM technology and shows significant industrial benefits.

Yuhua Jiang, Qihan Liu, Yiqin Yang et al.

Exploration in sparse reward environments remains a significant challenge in reinforcement learning, particularly in Contextual Markov Decision Processes (CMDPs), where environments differ across episodes. Existing episodic intrinsic motivation methods for CMDPs primarily rely on count-based approaches, which are ineffective in large state spaces, or on similarity-based methods that lack appropriate metrics for state comparison. To address these shortcomings, we propose Episodic Novelty Through Temporal Distance (ETD), a novel approach that introduces temporal distance as a robust metric for state similarity and intrinsic reward computation. By employing contrastive learning, ETD accurately estimates temporal distances and derives intrinsic rewards based on the novelty of states within the current episode. Extensive experiments on various benchmark tasks demonstrate that ETD significantly outperforms state-of-the-art methods, highlighting its effectiveness in enhancing exploration in sparse reward CMDPs.

Ni Mu, Hao Hu, Xiao Hu et al.

Preference-based reinforcement learning (PbRL) bypasses explicit reward engineering by inferring reward functions from human preference comparisons, enabling better alignment with human intentions. However, humans often struggle to label a clear preference between similar segments, reducing label efficiency and limiting PbRL's real-world applicability. To address this, we propose an offline PbRL method: Contrastive LeArning for ResolvIng Ambiguous Feedback (CLARIFY), which learns a trajectory embedding space that incorporates preference information, ensuring clearly distinguished segments are spaced apart, thus facilitating the selection of more unambiguous queries. Extensive experiments demonstrate that CLARIFY outperforms baselines in both non-ideal teachers and real human feedback settings. Our approach not only selects more distinguished queries but also learns meaningful trajectory embeddings.

Yifan Feng, Hao Hu, Xingliang Hou et al.

Large language models (LLMs) have transformed various sectors, including education, finance, and medicine, by enhancing content generation and decision-making processes. However, their integration into the medical field is cautious due to hallucinations, instances where generated content deviates from factual accuracy, potentially leading to adverse outcomes. To address this, we introduce Hyper-RAG, a hypergraph-driven Retrieval-Augmented Generation method that comprehensively captures both pairwise and beyond-pairwise correlations in domain-specific knowledge, thereby mitigating hallucinations. Experiments on the NeurologyCrop dataset with six prominent LLMs demonstrated that Hyper-RAG improves accuracy by an average of 12.3% over direct LLM use and outperforms Graph RAG and Light RAG by 6.3% and 6.0%, respectively. Additionally, Hyper-RAG maintained stable performance with increasing query complexity, unlike existing methods which declined. Further validation across nine diverse datasets showed a 35.5% performance improvement over Light RAG using a selection-based assessment. The lightweight variant, Hyper-RAG-Lite, achieved twice the retrieval speed and a 3.3% performance boost compared with Light RAG. These results confirm Hyper-RAG's effectiveness in enhancing LLM reliability and reducing hallucinations, making it a robust solution for high-stakes applications like medical diagnostics.

Yuwei Lou, Hao Hu, Shaocong Ma et al.

With the evolution of generative AI, multi - agent systems leveraging large - language models(LLMs) have emerged as a powerful tool for complex tasks. However, these systems face challenges in quantifying agent performance and lack mechanisms to assess agent credibility. To address these issues, we introduce DRF, a dynamic reputation filtering framework. DRF constructs an interactive rating network to quantify agent performance, designs a reputation scoring mechanism to measure agent honesty and capability, and integrates an Upper Confidence Bound - based strategy to enhance agent selection efficiency. Experiments show that DRF significantly improves task completion quality and collaboration efficiency in logical reasoning and code - generation tasks, offering a new approach for multi - agent systems to handle large - scale tasks.

Yongzhi Qi, Jiaheng Yin, Jianshen Zhang et al.

In supply chain management, planning is a critical concept. The movement of physical products across different categories, from suppliers to warehouse management, to sales, and logistics transporting them to customers, entails the involvement of many entities. It covers various aspects such as demand forecasting, inventory management, sales operations, and replenishment. How to collect relevant data from an e-commerce platform's perspective, formulate long-term plans, and dynamically adjust them based on environmental changes, while ensuring interpretability, efficiency, and reliability, is a practical and challenging problem. In recent years, the development of AI technologies, especially the rapid progress of large language models, has provided new tools to address real-world issues. In this work, we construct a Supply Chain Planning Agent (SCPA) framework that can understand domain knowledge, comprehend the operator's needs, decompose tasks, leverage or create new tools, and return evidence-based planning reports. We deploy this framework in JD.com's real-world scenario, demonstrating the feasibility of LLM-agent applications in the supply chain. It effectively reduced labor and improved accuracy, stock availability, and other key metrics.

Zhijun Zeng, Youjia Zheng, Hao Hu et al.

Accurate and efficient simulation of wave equations is crucial in computational wave imaging applications, such as ultrasound computed tomography (USCT), which reconstructs tissue material properties from observed scattered waves. Traditional numerical solvers for wave equations are computationally intensive and often unstable, limiting their practical applications for quasi-real-time image reconstruction. Neural operators offer an innovative approach by accelerating PDE solving using neural networks; however, their effectiveness in realistic imaging is limited because existing datasets oversimplify real-world complexity. In this paper, we present OpenBreastUS, a large-scale wave equation dataset designed to bridge the gap between theoretical equations and practical imaging applications. OpenBreastUS includes 8,000 anatomically realistic human breast phantoms and over 16 million frequency-domain wave simulations using real USCT configurations. It enables a comprehensive benchmarking of popular neural operators for both forward simulation and inverse imaging tasks, allowing analysis of their performance, scalability, and generalization capabilities. By offering a realistic and extensive dataset, OpenBreastUS not only serves as a platform for developing innovative neural PDE solvers but also facilitates their deployment in real-world medical imaging problems. For the first time, we demonstrate efficient in vivo imaging of the human breast using neural operator solvers.

Haoang Lu, Yuanqi Su, Xiaoning Zhang et al.

In recent years, visual 3D Semantic Scene Completion (SSC) has emerged as a critical perception task for autonomous driving due to its ability to infer complete 3D scene layouts and semantics from single 2D images. However, in real-world traffic scenarios, a significant portion of the scene remains occluded or outside the camera's field of view -- a fundamental challenge that existing monocular SSC methods fail to address adequately. To overcome these limitations, we propose Creating the Future SSC (CF-SSC), a novel temporal SSC framework that leverages pseudo-future frame prediction to expand the model's effective perceptual range. Our approach combines poses and depths to establish accurate 3D correspondences, enabling geometrically-consistent fusion of past, present, and predicted future frames in 3D space. Unlike conventional methods that rely on simple feature stacking, our 3D-aware architecture achieves more robust scene completion by explicitly modeling spatial-temporal relationships. Comprehensive experiments on SemanticKITTI and SSCBench-KITTI-360 benchmarks demonstrate state-of-the-art performance, validating the effectiveness of our approach, highlighting our method's ability to improve occlusion reasoning and 3D scene completion accuracy.

Hao Hu, Xuan Wang, Jingxiang Sun et al.

High-fidelity and efficient audio-driven talking head generation has been a key research topic in computer graphics and computer vision. In this work, we study vector image based audio-driven talking head generation. Compared with directly animating the raster image that most widely used in existing works, vector image enjoys its excellent scalability being used for many applications. There are two main challenges for vector image based talking head generation: the high-quality vector image reconstruction w.r.t. the source portrait image and the vivid animation w.r.t. the audio signal. To address these, we propose a novel scalable vector graphic reconstruction and animation method, dubbed VectorTalker. Specifically, for the highfidelity reconstruction, VectorTalker hierarchically reconstructs the vector image in a coarse-to-fine manner. For the vivid audio-driven facial animation, we propose to use facial landmarks as intermediate motion representation and propose an efficient landmark-driven vector image deformation module. Our approach can handle various styles of portrait images within a unified framework, including Japanese manga, cartoon, and photorealistic images. We conduct extensive quantitative and qualitative evaluations and the experimental results demonstrate the superiority of VectorTalker in both vector graphic reconstruction and audio-driven animation.

Wen Wang, Xiangchen Wu, Liang Wang et al.

This study addresses the Min-Max Multiple Traveling Salesmen Problem ($m^3$-TSP), which aims to coordinate tours for multiple salesmen such that the length of the longest tour is minimized. Due to its NP-hard nature, exact solvers become impractical under the assumption that $P \ne NP$. As a result, learning-based approaches have gained traction for their ability to rapidly generate high-quality approximate solutions. Among these, two-stage methods combine learning-based components with classical solvers, simplifying the learning objective. However, this decoupling often disrupts consistent optimization, potentially degrading solution quality. To address this issue, we propose a novel two-stage framework named \textbf{Generate-and-Split} (GaS), which integrates reinforcement learning (RL) with an optimal splitting algorithm in a joint training process. The splitting algorithm offers near-linear scalability with respect to the number of cities and guarantees optimal splitting in Euclidean space for any given path. To facilitate the joint optimization of the RL component with the algorithm, we adopt an LSTM-enhanced model architecture to address partial observability. Extensive experiments show that the proposed GaS framework significantly outperforms existing learning-based approaches in both solution quality and transferability.

Zhijun Zeng, Youjia Zheng, Chang Su et al.

Tissue mechanics--stiffness, density and impedance contrast--are broadly informative biomarkers across diseases, yet routine CT, MRI, and B-mode ultrasound rarely quantify them directly. While ultrasound tomography (UT) is intrinsically suited to in-vivo biomechanical assessment by capturing transmitted and reflected wavefields, efficient and accurate full-wave scattering models remain a bottleneck. Here, we introduce a generative neural physics framework that fuses generative models with physics-informed partial differential equation (PDE) solvers to produce rapid, high-fidelity 3D quantitative imaging of tissue mechanics. A compact neural surrogate for full-wave propagation is trained on limited cross-modality data, preserving physical accuracy while enabling efficient inversion. This enables, for the first time, accurate and efficient quantitative volumetric imaging of in vivo human breast and musculoskeletal tissues in under ten minutes, providing spatial maps of tissue mechanical properties not available from conventional reflection-mode or standard UT reconstructions. The resulting images reveal biomechanical features in bone, muscle, fat, and glandular tissues, maintaining structural resolution comparable to 3T MRI while providing substantially greater sensitivity to disease-related tissue mechanics.

Hao Hu, Xinqi Wang, Simon Shaolei Du

We introduce a novel task of clustering trajectories from offline reinforcement learning (RL) datasets, where each cluster center represents the policy that generated its trajectories. By leveraging the connection between the KL-divergence of offline trajectory distributions and a mixture of policy-induced distributions, we formulate a natural clustering objective. To solve this, we propose Policy-Guided K-means (PG-Kmeans) and Centroid-Attracted Autoencoder (CAAE). PG-Kmeans iteratively trains behavior cloning (BC) policies and assigns trajectories based on policy generation probabilities, while CAAE resembles the VQ-VAE framework by guiding the latent representations of trajectories toward the vicinity of specific codebook entries to achieve clustering. Theoretically, we prove the finite-step convergence of PG-Kmeans and identify a key challenge in offline trajectory clustering: the inherent ambiguity of optimal solutions due to policy-induced conflicts, which can result in multiple equally valid but structurally distinct clusterings. Experimentally, we validate our methods on the widely used D4RL dataset and custom GridWorld environments. Our results show that both PG-Kmeans and CAAE effectively partition trajectories into meaningful clusters. They offer a promising framework for policy-based trajectory clustering, with broad applications in offline RL and beyond.

Yiqin Yang, Quanwei Wang, Chenghao Li et al.

Offline reinforcement learning (RL) represents a significant shift in RL research, allowing agents to learn from pre-collected datasets without further interaction with the environment. A key, yet underexplored, challenge in offline RL is selecting an optimal subset of the offline dataset that enhances both algorithm performance and training efficiency. Reducing dataset size can also reveal the minimal data requirements necessary for solving similar problems. In response to this challenge, we introduce ReDOR (Reduced Datasets for Offline RL), a method that frames dataset selection as a gradient approximation optimization problem. We demonstrate that the widely used actor-critic framework in RL can be reformulated as a submodular optimization objective, enabling efficient subset selection. To achieve this, we adapt orthogonal matching pursuit (OMP), incorporating several novel modifications tailored for offline RL. Our experimental results show that the data subsets identified by ReDOR not only boost algorithm performance but also do so with significantly lower computational complexity.

Rui Yang, Yong Lin, Xiaoteng Ma et al.

Offline goal-conditioned RL (GCRL) offers a way to train general-purpose agents from fully offline datasets. In addition to being conservative within the dataset, the generalization ability to achieve unseen goals is another fundamental challenge for offline GCRL. However, to the best of our knowledge, this problem has not been well studied yet. In this paper, we study out-of-distribution (OOD) generalization of offline GCRL both theoretically and empirically to identify factors that are important. In a number of experiments, we observe that weighted imitation learning enjoys better generalization than pessimism-based offline RL method. Based on this insight, we derive a theory for OOD generalization, which characterizes several important design choices. We then propose a new offline GCRL method, Generalizable Offline goAl-condiTioned RL (GOAT), by combining the findings from our theoretical and empirical studies. On a new benchmark containing 9 independent identically distributed (IID) tasks and 17 OOD tasks, GOAT outperforms current state-of-the-art methods by a large margin.

Zhihan Liu, Miao Lu, Wei Xiong et al.

In online reinforcement learning (online RL), balancing exploration and exploitation is crucial for finding an optimal policy in a sample-efficient way. To achieve this, existing sample-efficient online RL algorithms typically consist of three components: estimation, planning, and exploration. However, in order to cope with general function approximators, most of them involve impractical algorithmic components to incentivize exploration, such as optimization within data-dependent level-sets or complicated sampling procedures. To address this challenge, we propose an easy-to-implement RL framework called \textit{Maximize to Explore} (\texttt{MEX}), which only needs to optimize \emph{unconstrainedly} a single objective that integrates the estimation and planning components while balancing exploration and exploitation automatically. Theoretically, we prove that \texttt{MEX} achieves a sublinear regret with general function approximations for Markov decision processes (MDP) and is further extendable to two-player zero-sum Markov games (MG). Meanwhile, we adapt deep RL baselines to design practical versions of \texttt{MEX}, in both model-free and model-based manners, which can outperform baselines by a stable margin in various MuJoCo environments with sparse rewards. Compared with existing sample-efficient online RL algorithms with general function approximations, \texttt{MEX} achieves similar sample efficiency while enjoying a lower computational cost and is more compatible with modern deep RL methods.

Xiaoteng Ma, Yiqin Yang, Hao Hu et al.

Offline reinforcement learning (RL) shows promise of applying RL to real-world problems by effectively utilizing previously collected data. Most existing offline RL algorithms use regularization or constraints to suppress extrapolation error for actions outside the dataset. In this paper, we adopt a different framework, which learns the V-function instead of the Q-function to naturally keep the learning procedure within the support of an offline dataset. To enable effective generalization while maintaining proper conservatism in offline learning, we propose Expectile V-Learning (EVL), which smoothly interpolates between the optimal value learning and behavior cloning. Further, we introduce implicit planning along offline trajectories to enhance learned V-values and accelerate convergence. Together, we present a new offline method called Value-based Episodic Memory (VEM). We provide theoretical analysis for the convergence properties of our proposed VEM method, and empirical results in the D4RL benchmark show that our method achieves superior performance in most tasks, particularly in sparse-reward tasks.

Zhizhou Ren, Guangxiang Zhu, Hao Hu et al.

Double Q-learning is a classical method for reducing overestimation bias, which is caused by taking maximum estimated values in the Bellman operation. Its variants in the deep Q-learning paradigm have shown great promise in producing reliable value prediction and improving learning performance. However, as shown by prior work, double Q-learning is not fully unbiased and suffers from underestimation bias. In this paper, we show that such underestimation bias may lead to multiple non-optimal fixed points under an approximate Bellman operator. To address the concerns of converging to non-optimal stationary solutions, we propose a simple but effective approach as a partial fix for the underestimation bias in double Q-learning. This approach leverages an approximate dynamic programming to bound the target value. We extensively evaluate our proposed method in the Atari benchmark tasks and demonstrate its significant improvement over baseline algorithms.

Hao Hu, Jianing Ye, Guangxiang Zhu et al.

Episodic memory-based methods can rapidly latch onto past successful strategies by a non-parametric memory and improve sample efficiency of traditional reinforcement learning. However, little effort is put into the continuous domain, where a state is never visited twice, and previous episodic methods fail to efficiently aggregate experience across trajectories. To address this problem, we propose Generalizable Episodic Memory (GEM), which effectively organizes the state-action values of episodic memory in a generalizable manner and supports implicit planning on memorized trajectories. GEM utilizes a double estimator to reduce the overestimation bias induced by value propagation in the planning process. Empirical evaluation shows that our method significantly outperforms existing trajectory-based methods on various MuJoCo continuous control tasks. To further show the general applicability, we evaluate our method on Atari games with discrete action space, which also shows a significant improvement over baseline algorithms.

Jin Zhang, Jianhao Wang, Hao Hu et al.

Meta reinforcement learning (meta-RL) extracts knowledge from previous tasks and achieves fast adaptation to new tasks. Despite recent progress, efficient exploration in meta-RL remains a key challenge in sparse-reward tasks, as it requires quickly finding informative task-relevant experiences in both meta-training and adaptation. To address this challenge, we explicitly model an exploration policy learning problem for meta-RL, which is separated from exploitation policy learning, and introduce a novel empowerment-driven exploration objective, which aims to maximize information gain for task identification. We derive a corresponding intrinsic reward and develop a new off-policy meta-RL framework, which efficiently learns separate context-aware exploration and exploitation policies by sharing the knowledge of task inference. Experimental evaluation shows that our meta-RL method significantly outperforms state-of-the-art baselines on various sparse-reward MuJoCo locomotion tasks and more complex sparse-reward Meta-World tasks.

Chunhua Jia, Lei Zhang, Hui Huang et al.

Multi-label networks with branches are proved to perform well in both accuracy and speed, but lacks flexibility in providing dynamic extension onto new labels due to the low efficiency of re-work on annotating and training. For multi-label classification task, to cover new labels we need to annotate not only newly collected images, but also the previous whole dataset to check presence of these new labels. Also training on whole re-annotated dataset costs much time. In order to recognize new labels more effectively and accurately, we propose GraftNet, which is a customizable tree-like network with its trunk pretrained with a dynamic graph for generic feature extraction, and branches separately trained on sub-datasets with single label to improve accuracy. GraftNet could reduce cost, increase flexibility, and incrementally handle new labels. Experimental results show that it has good performance on our human attributes recognition task, which is fine-grained multi-label classification.

Bo wang, Hao Hu, Caixia Zhang

Traditionally the danger cylinder is intimately related to the solution stability in P3P problem. In this work, we show that the danger cylinder is also closely related to the multiple-solution phenomenon. More specifically, we show when the optical center lies on the danger cylinder, of the 3 possible P3P solutions, i.e., one double solution, and two other solutions, the optical center of the double solution still lies on the danger cylinder, but the optical centers of the other two solutions no longer lie on the danger cylinder. And when the optical center moves on the danger cylinder, accordingly the optical centers of the two other solutions of the corresponding P3P problem form a new surface, characterized by a polynomial equation of degree 12 in the optical center coordinates, called the Companion Surface of Danger Cylinder (CSDC). That means the danger cylinder always has a companion surface. For the significance of CSDC, we show that when the optical center passes through the CSDC, the number of solutions of P3P problem must change by 2. That means CSDC acts as a delimitating surface of the P3P solution space. These new findings shed some new lights on the P3P multi-solution phenomenon, an important issue in PnP study.