Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Ho-fung LeungSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works382 from public data
- Application of artificial intelligence to wastewater treatment: A bibliometric analysis and systematic review of technology, economy, management, and wastewater reuse440
This study examines the literature from 1995 to 2019 to conduct a large-scale bibliometric analysis of trends in the application of artificial intelligence technology to wastewater treatment to provide perspectives on the potential future directions of new research frontiers in the utilization of artificial Intelligence in wastewater treatment plants that simultaneously address pollutant removal, cost reduction, water reuse, and management challenges in complex practical applications.
- 426
- A Survey on Vision--Language--Action Models for Embodied AI403
This work provides a detailed taxonomy of VLAs, organized into three major lines of research, dedicated to developing VLA-based control policies adept at predicting low-level actions and guiding VLAs to follow more general user instructions.
- Multi-Agent Game Abstraction via Graph Attention Neural Network355
A novel game abstraction mechanism based on two-stage attention network (G2ANet) is proposed, which can indicate whether there is an interaction between two agents and the importance of the interaction and two novel learning algorithms GA-Comm and GA-AC are proposed.
- Exploration in Deep Reinforcement Learning: From Single-Agent to Multiagent Domain207
This article provides a comprehensive and unified empirical comparison of different exploration methods for DRL on a set of commonly used benchmarks and summarizes the open problems of exploration in DRL and deep MARL and point out a few future directions.
- 199
- A survey on interpretable reinforcement learning189
This survey provides an overview of various approaches to achieve higher interpretability in reinforcement learning (RL), and argues that interpretable RL may embrace different facets: interpretable inputs, interpretable (transition/reward) models, and interpretable decision-making.
- An end-to-end heterogeneous graph representation learning-based framework for drug–target interaction prediction179
A novel 'end-to-end' learning-based framework based on heterogeneous 'graph' convolutional networks for 'DTI' prediction called end- to-end graph (EEG)-DTI, which performs better than existing state-of-art methods for identifying DTIs.
- A learning-based framework for miRNA-disease association identification using neural networks178
A novel learning-based framework, MDA-CNN, is proposed, which outperforms some state-of-the-art approaches in a large margin on both tasks of miRNA-disease association prediction and mi RNA-phenotype association prediction.
- From Few to More: Large-Scale Dynamic Multiagent Curriculum Learning147
A novel Dynamic Multiagent Curriculum Learning (DyMA-CL) to solve large-scale problems by starting from learning on a multiagent scenario with a small size and progressively increasing the number of agents, and proposes three transfer mechanisms across curricula to accelerate the learning process.
- Modeling Scale-free Graphs with Hyperbolic Geometry for Knowledge-aware Recommendation117
LKGR facilitates better modeling of scale-free tripartite graphs after the data unification and employs different information propagation strategies in the hyperbolic space to explicitly encode heterogeneous information from historical interactions and KGs.
- 114
- Falsification of Cyber-Physical Systems Using Deep Reinforcement Learning111
State-of-the-art Deep Reinforcement Learning techniques are explored, i.e., Asynchronous Advantage Actor-Critic and Double Deep Q Network (DDQN), to reduce the number of simulation runs required to find such counterexamples and identify two factors of CPS which make DRL based methods better than existing methods.
- Bridging Evolutionary Algorithms and Reinforcement Learning: A Comprehensive Survey on Hybrid Algorithms108
This survey offers a comprehensive overview of the diverse research branches in ERL, systematically summarize the recent advancements in related algorithms and identify three primary research directions: 1) EA-assisted optimization of RL; 2) RL-assisted optimization of EA; and 3) synergistic optimization of EA and RL.
- 105
- Automatic Web Testing Using Curiosity-Driven Reinforcement Learning104
This paper proposes WebExplor, an automatic end-to-end web testing framework that adopts a curiosity-driven reinforcement learning to generate high-quality action sequences (test cases) with temporal logical relations that can achieve significantly higher failure detection rate, code coverage and efficiency than existing state-of-the-art web testing techniques.
- CellAgent: An LLM-driven Multi-Agent Framework for Automated Single-cell Data Analysis98
This work introduces CellAgent, an LLM-driven multi-agent framework, specifically designed for the automatic processing and execution of scRNA-seq data analysis tasks, providing high-quality results with no human intervention.
- Learning to select cuts for efficient mixed-integer programming98
A data-driven and generalizable cut selection approach, named Cut Ranking, in the settings of multiple instance learning, which has been deployed in an industrial solver for large-scale MIPs and has achieved the average speedup ratio of 12.42% over the production solver without any accuracy loss of solution.
- Ordering-Based Causal Discovery with Reinforcement Learning96
This work forms the ordering search problem as a multi-step Markov decision process, implements the ordering generating process with an encoder-decoder architecture, and uses RL to optimize the proposed model based on the reward mechanisms designed for each ordering.
- 96
- 96
- 94
- A benchmark for automatic medical consultation system: frameworks, tasks and datasets93
This article creates a new large medical dialogue dataset with multi-level fine-grained annotations and establishes five independent tasks, including named entity recognition, dialogue act classification, symptom label inference, medical report generation and diagnosis-oriented dialogue policy.
- SPA-Bench: A Comprehensive Benchmark for SmartPhone Agent Evaluation90
SPA-Bench is presented, a comprehensive SmartPhone Agent Benchmark designed to evaluate (M)LLM-based agents in an interactive environment that simulates real-world conditions and proposes future research directions to ease difficulties, moving closer to real-world smartphone agent applications.
- Deep Multi-Agent Reinforcement Learning with Discrete-Continuous Hybrid Action Spaces89
The empirical results show that both Deep MAPQN and Deep MAHHQN are effective and significantly outperform existing independent deep parameterized Q-learning method.
- ChiPFormer: Transferable Chip Placement via Offline Decision Transformer88
This work casts the chip placement as an offline RL formulation and presents ChiPFormer that enables learning a transferable placement policy from fixed offline data and achieves significantly better placement quality while reducing the runtime by 10x compared to recent state-of-the-art approaches.
- SC2disease: a manually curated database of single-cell transcriptome for human diseases88
The literature prior to March 2020 for studies which used scRNA-seq to study diseases with human samples are reviewed, and the SC2disease database is developed to summarize all the data by different diseases, tissues and cell types.
- Neighborhood Cognition Consistent Multi-Agent Reinforcement Learning87
This work proposes neighborhood cognition consistent deep Q-learning and Actor-Critic to facilitate large-scale multi-agent cooperations and justifies the superior performance of the methods compared with state-of-the-art MARL approaches.
- MetaDiffuser: Diffusion Model as Conditional Planner for Offline Meta-RL85
A task-oriented conditioned diffusion planner for offline meta-RL(MetaDiffuser), which considers the generalization problem as conditional trajectory generation task with contextual representation, and designs a dual-guided module in the sampling process of the diffusion model.
- HEBO: An Empirical Study of Assumptions in Bayesian Optimisation83
The findings indicate that the majority of hyper-parameter tuning tasks exhibit heteroscedasticity and non-stationarity, multiobjective acquisition ensembles with Pareto front solutions improve queried configurations, and robust acquisition maximisers afford empirical advantages relative to their non-robust counterparts.
- A Deep Bayesian Policy Reuse Approach Against Non-Stationary Agents83
A new deep BPR+ algorithm is proposed by extending the recent BPR+, with a neural network as the value-function approximator and taking advantage of therectified belief model to detect policy accurately and infer the other agent's policy from reward signals and its behaviors.
- Traj-MAE: Masked Autoencoders for Trajectory Prediction80
This work proposes an efficient masked autoencoder for trajectory prediction (Traj-MAE) that better represents the complicated behaviors of agents in the driving environment and introduces a continual pre-training framework that can help Traj-MAE learn valuable and diverse information from various strategies efficiently.
- Attention-based recurrent neural network for influenza epidemic prediction74
A new deep neural network structure is proposed that forecasts a real-time influenza-like illness rate (ILI%) in Guangzhou, China using long short-term memory (LSTM) neural networks to precisely forecast accurateness due to the long-term attribute and diversity of influenza epidemic data.
- SpatialCoT: Advancing Spatial Reasoning through Coordinate Alignment and Chain-of-Thought for Embodied Task Planning73
This work proposes a novel approach named SpatialCoT, specifically designed to bolster the spatial reasoning capabilities of Vision-Language Models (VLMs), which significantly outperforms previous state-of-the-art approaches in both navigation and manipulation tasks.
- 71
- 67
- Out-of-distribution Detection with Implicit Outlier Transformation65
This work proposes a novel OE-based approach that makes the model perform well for unseen OOD situations, even forseen OOD cases, and offers an efficient way of synthesizing OOD data, which can further benefit the detection model, besides the surrogate Ood data.
- Towards Effective Context for Meta-Reinforcement Learning: an Approach based on Contrastive Learning63
This work argues that improving the quality of context involves answering two questions: how to train a compact and sufficient encoder that can embed the task-specific information contained in prior trajectories and how to collect informative trajectories of which the corresponding context reflects the specification of tasks.
- A Multi-Graph Attributed Reinforcement Learning based Optimization Algorithm for Large-scale Hybrid Flow Shop Scheduling Problem61
A novel Multi-Graph Attributed Reinforcement Learning based Optimization (MGRO) algorithm is proposed to better tackle the practical large-scale HFSP and improve the existing algorithm.
- PMIC: Improving Multi-Agent Reinforcement Learning with Progressive Mutual Information Collaboration59
A novel MARL framework, called Progressive Mutual Information Collaboration (PMIC), is proposed, which uses a new collaboration criterion measured by the MI between global states and joint actions and plays complementary roles by facilitating better collaborations while avoiding falling into sub-optimal ones.
- A robot operating system framework for using large language models in embodied AI56
A framework for intuitive robot programming by non-experts, leveraging natural language prompts and contextual information from the Robot Operating System (ROS), enabling non-experts to articulate task requirements to the system through a chat interface is presented.
- Contrastive-ACE: Domain Generalization Through Alignment of Causal Mechanisms52
This work considers the invariance of the average causal effect of the features to the labels to enforce stability of the causal prediction by the classifier across domains, and sheds some light on the domain generalization problem by introducing invariances of the mechanisms into the learning process.
- A Graph-Enhanced Click Model for Web Search51
A novel graph-enhanced click model (GraphCM) for web search that not only outperforms the state-of-art models, but also achieves superior performance in addressing the data sparsity and cold-start problems.
- Generating Behavior-Diverse Game AIs with Evolutionary Multi-Objective Deep Reinforcement Learning50
Evaluations on the Atari and real commercial games indicate that, compared to existing algorithms, EMOGI performs better in generating diverse behaviors and significantly improves the efficiency of Game AI design.
- Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulation49
This work shows that a pointing-centric representation, combined with an RFT training paradigm, offers an effective and generalizable pathway to closing the perception-action gap in robotics.
- 49
- Efficient Deep Reinforcement Learning via Adaptive Policy Transfer48
A novel Policy Transfer Framework (PTF) is proposed that learns when and which source policy is the best to reuse for the target policy and when to terminate it by modeling multi-policy transfer as an option learning problem.
- The dynamics of reinforcement social learning in networked cooperative multiagent systems48
This work systematically investigates the multiagent coordination problems in cooperative environments under the networked social learning framework under four representative topologies and proposes two types of learners: individual action learner and joint action learners.
- Towards Efficient Detection and Optimal Response against Sophisticated Opponents47
A novel approach called Bayes-ToMoP which can efficiently detect the strategy of opponents using either stationary or higher-level reasoning strategies, and also supports the detection of previously unseen policies and learning a best-response policy accordingly.
- AlignDiff: Aligning Diverse Human Preferences via Behavior-Customisable Diffusion Model45
This work proposes AlignDiff, a novel framework that leverages RL from Human Feedback (RLHF) to quantify human preferences, covering abstractness, and utilizes them to guide diffusion planning for zero-shot behavior customizing, covering mutability.
- ERL-Re$^2$: Efficient Evolutionary Reinforcement Learning with Shared State Representation and Individual Policy Representation44
Evolutionary Reinforcement Learning with Two-scale State Representation and Policy Representation (ERL-Re$^2$), a novel solution to the aforementioned two drawbacks of Deep RL, is proposed.
- KoGuN: Accelerating Deep Reinforcement Learning via Integrating Human Suboptimal Knowledge42
This work proposes knowledge guided policy network (KoGuN), a novel framework that combines human prior suboptimal knowledge with reinforcement learning and achieves significant improvement on learning efficiency of flat RL algorithms, even with very low-performance human prior knowledge.
- 41
- 41
- 39
- Weighted Double Deep Multiagent Reinforcement Learning in Stochastic Cooperative Environments39
This work applies a specific adaptation to one recently proposed weighted double estimator and proposes a multiagent deep reinforcement learning framework, named Weighted Double Deep Q-Network (WDDQN), which outperforms the existing DRL algorithm (double DQN) and multiagent RL algorithm (lenient Q-learning) in terms of performance and convergence within stochastic cooperative environments.
- 37
- DAG Matters! GFlowNets Enhanced Explainer For Graph Neural Networks36
The GFlowExplainer aims to learn a policy that generates a distribution of subgraphs for which the probability of a subgraph is proportional to its' reward, and proposes a new cut vertex matrix to efficiently explore parent states for GFlowNets structure, thus making the approach applicable in a large-scale setting.
- Triple-GAIL: A Multi-Modal Imitation Learning Framework with Generative Adversarial Nets36
A novel multi-modal GAIL framework, named Triple-GAIL, that is able to learn skill selection and imitation jointly from both expert demonstrations and continuously generated experiences with data augmentation purpose by introducing an auxiliary selector is proposed.
- Recurrent Deep Multiagent Q-Learning for Autonomous Brokers in Smart Grid36
This paper develops an effective pricing strategy for brokers in local electricity retail market based on recurrent deep multiagent reinforcement learning and sequential clustering and highlights the effectiveness of the reward shaping mechanism.
- Spatial-Temporal Graph Diffusion Policy with Kinematic Modeling for Bimanual Robotic Manipulation35
Experimental results show that the proposed Kinematics enhanced Spatial-TemporAl gRaph Diffuser effectively leverages the physical structural information and generates kinematics-aware actions in both simulation and real-world.
- Mem2Ego: Empowering Vision-Language Models with Global-to-Ego Memory for Long-Horizon Embodied Navigation34
A novel vision-language model (VLM)-based navigation framework that addresses challenges by adaptively retrieving task-relevant cues from a global memory module and integrating them with the agent's egocentric observations and enhances spatial reasoning and decision-making in long-horizon tasks.
- SheetAgent: Towards a Generalist Agent for Spreadsheet Reasoning and Manipulation via Large Language Models34
This work introduces SheetRM, a benchmark featuring long-horizon and multi-category tasks with reasoning-dependent manipulation caused by real-life challenges, and proposes SheetAgent, a novel autonomous agent that utilizes the power of LLMs.
- Large-Scale Home Energy Management Using Entropy-Based Collective Multiagent Deep Reinforcement Learning Framework33
This paper focuses on a microgrid in which a large-scale modern homes interact together to optimize their electricity cost, and presents an Entropy-Based Collective Multiagent Deep Reinforcement Learning (EB-C-MADRL) framework to address it.
- Pangu-Agent: A Fine-Tunable Generalist Agent with Structured Reasoning32
A general framework model for integrating and learning structured reasoning into AI agents' policies and indicates that AI agents perform and adapt far better when organised reasoning and prior knowledge are embedded, opening the door to more resilient and general AI agent systems.
- SEIHAI: A Sample-Efficient Hierarchical AI for the MineRL Competition32
This paper split the task into several sequentially dependent subtasks, and train a suitable agent for each subtask using reinforcement learning and imitation learning, and design a scheduler to select different agents for different subtasks automatically.
- 32
- 32
- 32
- Addressing Action Oscillations through Learning Policy Inertia31
Policy Inertia Controller (PIC) is introduced which serves as a generic plug-in framework to off-the-shelf DRL algorithms, to enable adaptive balance between the optimality and smoothness in a formal way and Nested Policy Iteration is proposed as a general training algorithm for PIC-augmented policy which ensures monotonically non-decreasing updates.
- Uni-RLHF: Universal Platform and Benchmark Suite for Reinforcement Learning with Diverse Human Feedback30
Uni-RLHF develops a user-friendly annotation interface tailored to various feedback types, compatible with a wide range of mainstream RL environments, and establishes a systematic pipeline of crowdsourced annotations, resulting in large-scale annotated datasets comprising more than 15 million steps across 30+ popular tasks.
- 30
- DyPS: Dynamic Parameter Sharing in Multi-Agent Reinforcement Learning for Spatio-Temporal Resource Allocation27
This paper designs self-supervised learning tasks to extract the implicit behavioral characteristics from the action trajectories of agents and proposes a hierarchical MARL framework capable of dynamically revising the agent role divisions during the training process and thus shares parameters among agents with the same role, reducing computational complexity.
- Learnable Behavior Control: Breaking Atari Human World Records via Sample-Efficient Behavior Selection27
A general framework called Learnable Behavioral Control (LBC) is proposed to address the limitation and enables a significantly enlarged behavior selection space via formulating a hybrid behavior mapping from all policies; and constructs a unified learnable process for behavior selection.
- An efficient and robust negotiating strategy in bilateral negotiations over multiple items27
Through analysis, the ABiNeS strategy is shown to be very robust against other state-of-the-art strategies under different negotiation contexts, and adopted the single-agent best deviation principle to analyze the robustness of different negotiation strategies based on model checking techniques.
- 26
- Generalizable Floorplanner through Corner Block List Representation and Hypergraph Embedding26
This work proposes a novel deep reinforcement learning agent to perform floorplanning, one of the early stages of VLSI physical design, and is the first work to select both id and block position with an entirely end-to-end learning-based framework that can generalize.
- 26
- A Method for Deploying Distributed Denial of Service Attack Defense Strategies on Edge Servers Using Reinforcement Learning26
A novel deep reinforcement learning framework is developed to balance the deployment of computing resource and tasks allocation, in which graph neural network used to coding the network structure information transformation as vector, and the traffic information to input into Q-Network to obtain the best allocation results.
- Blind Image Denoising via Dependent Dirichlet Process Tree26
This paper proposes a novel blind image denoising algorithm to recover the clean image from noisy one with the unknown noise model, and derives a variational inference algorithm to estimate model parameters and recover clean patches.
- Apollo-MILP: An Alternating Prediction-Correction Neural Solving Framework for Mixed-Integer Linear Programming25
This work proposes an Alternating prediction-correction neural solving framework (Apollo-MILP) that can identify and select accurate and reliable predicted values to fix that significantly outperforms other ML-based approaches in terms of solution quality, achieving over a 50% reduction in the solution gap.
- Generalized Universal Domain Adaptation with Generative Flow Networks25
This work introduces a new problem in unsupervised domain adaptation, termed as Generalized Universal Domain Adaptation (GUDA), which aims to achieve precise prediction of all target labels including unknown categories, and proposes an active domain adaptation algorithm named GFlowDA, which selects diverse samples with probabilities proportional to a reward function.
- CFlowNets: Continuous Control with Generative Flow Networks25
Experimental results on continuous control tasks demonstrate the performance advantages of CFlowNets compared to many reinforcement learning methods, especially regarding exploration ability.
- ProRefiner: an entropy-based refining strategy for inverse protein folding with global graph attention25
This work proposes an entropy-based residue selection method to remove noise in the input residue context, and introduces ProRefiner, a memory-efficient global graph attention model to fully utilize the denoised context.
- Multiagent Reinforcement Social Learning toward Coordination in Cooperative Multiagent Systems25
This work investigates the multiagent coordination problems in cooperative environments under a social learning framework, and distinguishes two different types of learners depending on the amount of information each agent can perceive: individual action learner and joint action learners.
- ViMo: A Generative Visual GUI World Model for App Agents24
This work proposes ViMo, the first visual world model designed to generate future App observations as images and proposes a novel data representation, the Symbolic Text Representation~(STR) to overlay text content with symbolic placeholders while preserving graphics.
- Breaking Filter Bubble: A Reinforcement Learning Framework of Controllable Recommender System24
The proposed framework of controllable recommendation in this work can inspire not only the researchers of recommender systems, but also a broader community concerned with artificial intelligence algorithms’ impact on humanity, especially for those vulnerable populations on the Web.
- Improved Sample Complexity Bounds for Branch-And-Cut24
These guarantees apply to parameters that control the most important aspects of branch-and-cut: node selection, branching constraint selection, and cutselection, and are sharper and more general than those from prior research.
- Background-free upconversion-encoded microspheres for mycotoxin detection based on a rapid visualization method24
A quantitative fluorescence image analysis based on multicolor upconversion nanocrystal (UCN)-encoded microspheres for detection of ochratoxin A and zearalenone is constructed and it is shown that this novel detection platform performed feasible and reliablefluorescence image measurements by this method.
- 24
- Hierarchical Deep Multiagent Reinforcement Learning24
This paper decomposes the original MARL problem into hierarchies and investigates how effective policies can be learned hierarchically in synchronous/asynchronous hierarchical MARL frameworks, and proposes a new experience replay mechanism, named as Augmented Concurrent Experience Replay (ACER).
- 24
- Uncertainty-aware Consistency Learning for Cold-Start Item Recommendation23
A Uncertainty-aware Consistency learning framework for Cold-start item recommendation (shorten as UCC) solely based on user-item interactions is proposed, which significantly outperforms state-of-the-art methods on both warm and cold items.
- Empirical Policy Optimization for n-Player Markov Games23
This research treats the evolution of player policies as a dynamical process and proposes a novel learning scheme for Nash equilibrium, which develops an empirical policy optimization algorithm which is implemented in a reinforcement-learning framework and runs in a distributed way, with each player optimizing its policy based on own observations.
- 23
- CUHKAgent: An Adaptive Negotiation Strategy for Bilateral Negotiations over Multiple Items23
An adaptive negotiating strategy (CUHKAgent) is proposed to predict the opponent’s strategy and preference at a high level, and make informed decision accordingly.
- SA-IGA: a multiagent reinforcement learning method towards socially optimal outcomes22
A novel gradient ascent based algorithm (SA-IGA) which augments the basic gradient-ascent algorithm by incorporating social awareness into the policy update process and shows that SA-PGA agent can achieve higher social welfare than previous social-optimality oriented Conditional Joint Action Learner and also is robust against individually rational opponents by reaching Nash equilibrium solutions.
- What about Inputting Policy in Value Function: Policy Representation and Policy-Extended Value Function Approximator21
It is shown that generalized value estimates offered by PeVFA may have lower initial approximation error to true values of successive policies, which is expected to improve consecutive value approximation during GPI.
- 21
- 21
- Efficient Convention Emergence through Decoupled Reinforcement Social Learning with Teacher-Student Mechanism21
This paper designs reinforcement learning based (RL-based) strategies to promote convention emergence in multiagent systems (MASs) with large convention space and proposes two learning strategies, multiple-Q and multiple-R, and proposes incorporating teacher-student mechanism on top of the learning strategies to accelerate lexicon convergence speed.
- AgentSwift: Efficient LLM Agent Design via Value-guided Hierarchical Search20
This work formalizes a hierarchical search space that jointly models agentic workflow and composable functional components, which enables the discovery of more complex and effective agent architectures and serves as a launchpad for researchers to rapidly prototype and discover powerful agent architectures without the impediment of prohibitive evaluation costs.
- Generative Models in Decision Making: A Survey20
This survey proposes a principled taxonomy grounding generative decision-making within the probabilistic framework of Control as Inference, and conceptualizes four distinct functional roles: controllers for amortized policy inference, Modelers for dynamics priors, Optimizers for iterative trajectory refinement, and Evaluators for trajectory guidance and value assessment.
- DeepRisk: A deep learning approach for genome-wide assessment of common disease risk20
DeepRisk is developed, a biological knowledge-driven deep learning method for modeling these complex, nonlinear associations among SNPs, to provide a more effective method for scoring the risk of common diseases with genome-wide genotype data.
- Learning Adaptive Display Exposure for Real-Time Advertising20
Experimental evaluations on industry-scale real-world datasets demonstrate the merits of the approach in both obtaining higher revenue under the constraints and the effectiveness of the constrained hindsight experience replay mechanism.
- Modelling Adaptive Learning Behaviours for Consensus Formation in Human Societies20
Experimental results show that the proposed model can facilitate the formation of consensus among agents, and some critical factors such as size of opinion space and network topology can have significant influences on opinion dynamics.
- ET-Plan-Bench: Embodied Task-level Planning Benchmark Towards Spatial-Temporal Cognition with Foundation Models19
This work introduces a novel embodied task planning benchmark, ET-Plan-Bench, designed to evaluate two critical dimensions of LLMs’ application in embodied task understanding: spatial understanding and temporal and causal comprehension of sequences of actions within an environment.
- STAR: Learning Diverse Robot Skill Abstractions through Rotation-Augmented Vector Quantization19
To prevent codebook collapse, rotation-augmented residual skill quantization (RaRSQ) is devised, which encodes relative angles between encoder outputs into the gradient flow by rotation-based gradient mechanism.
- Identifying potential risk genes for clear cell renal cell carcinoma with deep reinforcement learning19
A deep reinforcement learning-based computational approach named RL-GenRisk is proposed to identify ccRCC risk genes as a Markov decision process, combining the graph convolutional network and Deep Q-Network for risk gene identification.
- Enabling target-aware molecule generation to follow multi objectives with Pareto MCTS19
This article proposes ParetoDrug, a Pareto Monte Carlo Tree Search (MCTS) generation algorithm, which utilizes pretrained atom-by-atom autoregressive generative models for the exploration guidance to desired molecules during MCTS searching and is highly effective in traversing the large and complex chemical space.
- Learning from Hierarchical Structure of Knowledge Graph for Recommendation19
This work proposes to reveal unknown hierarchical structures with a supervised signal detection method and then exploit the hierarchical structure with disentangling representation learning, and conducts experiments on two large-scale datasets to verify the superiority and rationality of the proposed method.
- Continuous Multiagent Control Using Collective Behavior Entropy for Large-Scale Home Energy Management19
A collective MA-DRL algorithm with continuous action space to provide fine-grained control on a large scale microgrid to mitigate the non-stationarity of the microgrid environment, and a novel predictive model is proposed to measure the collective market behavior.
- 19
- 19
- Independent Generative Adversarial Self-Imitation Learning in Cooperative Multiagent Systems19
This work is the first to combine self-imitation learning with generative adversarial imitation learning (GAIL) and apply it to cooperative multiagent systems and produces state-of-the-art results and even outperforms JALs in terms of both convergence speed and final performance.
- 19
- Achieving Socially Optimal Outcomes in Multiagent Systems with Reinforcement Social Learning19
This work proposes an alternative social learning framework that is suitable for both symmetric and asymmetric games, and proposes that the agents’ learning strategies should be based on reinforcement learning technique instead of evolutionary learning.
- The rise and potential opportunities of large language model agents in bioinformatics and biomedicine18
The technical foundations of LLM agents, including their core architecture, key technologies, and collaborative modes, are reviewed, and the applications of LLM agents in multi-omics, drug development, chemical research, clinical diagnosis, and health management are explored.
- 18
- Achieving cooperation through deep multiagent reinforcement learning in sequential prisoner's dilemmas18
This work proposes a deep multiagent reinforcement-learning approach that investigates the evolution of mutual cooperation in SPD games and shows that its strategy can avoid being exploited by exploitative opponents and achieve cooperation with cooperative opponents.
- Efficient and Robust Emergence of Norms through Heuristic Collective Learning18
This article proposes two novel learning strategies under the collective learning framework, collective learning EV-l and collectivelearning EV-g, to efficiently facilitate the emergence of social norms and shows that both learning strategies can support the emerged of desirable social norms more efficiently.
- 18
- AppVLM: A Lightweight Vision Language Model for Online App Control17
This work introduces AppVLM, a lightweight Vision-Language Model (VLM), and indicates that AppVLM achieves the highest action prediction accuracy in offline evaluation on the AndroidControl dataset, compared to all evaluated baselines, and matches GPT-4o in online task completion success rate in the AndroidWorld environment, while being up to ten times faster.
- Towards Next-Generation Logic Synthesis: A Scalable Neural Circuit Generation Framework17
A novel regularized triangle-shaped circuit network generation framework, which leverages the key insights for completely accurate and scalable circuit generation and an evolutionary algorithm assisted by reinforcement learning agent restarting technique for efficient and effective neural circuit optimization.
- A Circuit Domain Generalization Framework for Efficient Logic Synthesis in Chip Design17
PruneX is the first approach to tackle the out-of-distribution (OOD) generalization problem in LS operators, and is integrated with the Resub and Mfs2 operators to reduce ineffective transformations.
- Debiased Recommendation with User Feature Balancing17
This article proposes a novel debiased recommendation framework based on user feature balancing, which aims to remove the causal relation from the user to the item, which enables us to achieve unbiased recommendation, bypassing the computation of IPS.
- 17
- Rethinking Decision Transformer via Hierarchical Reinforcement Learning17
This work introduces a general sequence modeling framework for studying sequential decision making through the lens of Hierarchical RL, and shows DT emerges as a special case of this framework with certain choices of high-level and low-level policies, and discusses the potential failure of these choices.
- SplitNet: A Reinforcement Learning Based Sequence Splitting Method for the MinMax Multiple Travelling Salesman Problem17
This paper generates single TSP solution sequences and split them into mTSP subsequences using an attention-based model trained by reinforcement learning and designs a decision region for the splitting policy, which significantly reduces the policy action space on instances of various scales and thus improves the generalization ability of SplitNet.
- EUCLID: Towards Efficient Unsupervised Reinforcement Learning with Multi-choice Dynamics Model17
An Efficient Unsupervised Reinforcement Learning Framework with Multi-choice Dynamics model (EUCLID) is proposed, which introduces a novel model-fused paradigm to jointly pre-train the dynamics model and unsupervised exploration policy in the pre-training phase, thus better leveraging the environmental samples and improving the downstream task sampling efficiency.
- 17
- Efficient policy detecting and reusing for non-stationarity in Markov games17
A new deep Bayesian policy reuse algorithm, a.k.a. DPN-BPR+ is proposed, by extending the recent BPR+ algorithm with a neural network as the value-function approximator and taking advantage of the opponent model to infer the other agents’ policy from reward signals and its behavior.
- Hierarchical Learning for Emergence of Social Norms in Networked Multiagent Systems17
Results verify that learning from local interactions integrating hierarchical supervision can be an effective mechanism for emergence of social norms in networked multiagent systems.
- Using deep reinforcement learning to speed up collective cell migration16
An agent-based modeling framework is used to establish a simulation platform for simulating collective cell migration and the goal of the platform is to build a biomimetic environment to demonstrate the importance of stimuli between the leading and following cells.
- Optimal Personalized Defense Strategy Against Man-In-The-Middle Attack16
A game-theoretic defense strategy is proposed from a different perspective, which aims at minimizing the loss that the whole system sustains given that the MITM attacks are inevitable.
- 16
- CoopRide: Cooperate All Grids in City-Scale Ride-Hailing Dispatching with Multi-Agent Reinforcement Learning15
The CoopRide framework is proposed, which uniformly encode both within- and cross-grid dispatching, enabling flexible choice of both within- and cross-grid dispatching and to automatically learn the cooperation intensities among grids, thereby obtaining the cooperative rewards to drive the learning of global cooperation actions.
- 15
- GFlowCausal: Generative Flow Networks for Causal Discovery15
A novel approach to learning a DAG from observational data called GFlowCausal, which converts the graph search problem to a generation problem, in which direct edges are added gradually and could guarantee acyclicity properties effectively and the consistency between final states and fully-connected graphs.
- Cell‐to‐cell and type‐to‐type heterogeneity of signaling networks: insights from the crowd15
The challenge results show that despite the stochastic nature of signal transduction in single cells, the signaling events are tightly controlled and machine learning methods can accurately predict new experimental data.
- 15
- 15
- 15
- 15
- Evaluating Practical Automated Negotiation Based on Spatial Evolutionary Game Theory15
It is shown how both empirical and spatial evolutionary game theory can be used to interpret bilateral negotiation results among state of the art negotiating agents in these different scenarios.
- 14
- Prototypical Context-Aware Dynamics for Generalization in Visual Control With Model-Based Reinforcement Learning14
This work proposes a prototypical context- aware dynamics (ProtoCAD) model, which captures local dynamics using temporally consistent latent contexts and aids generalization in visual control tasks.
- Spectral Augmentations for Graph Contrastive Learning14
A set of well-motivated graph transformation operations derived via graph spectral analysis is proposed to provide a bank of candidates when constructing augmentations for a graph contrastive objective, enabling contrastive learning to capture useful structural representation from pre-training graph datasets.
- Coach-assisted multi-agent reinforcement learning framework for unexpected crashed agents14
This work proposes a coach-assisted multi-agent reinforcement learning framework that introduces a virtual coach agent to adjust the crash rate during training and designs three coaching strategies (fixed crash rate, curriculum learning, and adaptive crash rate) and a re-sampling strategy for the authors' coach agent.
- Plan Your Target and Learn Your Skills: Transferable State-Only Imitation Learning via Decoupled Policy Optimization14
This paper introduces Decoupled Policy Optimization (DePO), which explicitly decouples the policy as a high-level state planner and an inverse dynamics model and demonstrates the appealing usage of DePO for transferring across different tasks by pre-training, and the potential for co-training agents with various skills.
- 14
- Automatic Feature Engineering by Deep Reinforcement Learning14
A Deep Q-learning on HTG is proposed to support efficient learning of fine-grained and generalized FE policies that can transfer knowledge of engineering "good" features from a collection of datasets to other unseen datasets.
- Accelerating Norm Emergence Through Hierarchical Heuristic Learning14
A hierarchically heuristic learning strategy (HHLS) under the hierarchical social learning framework is proposed that can support the emergence of desirable social norms more efficiently and can be applicable in a much wider range of multiagent interaction scenarios compared with previous work.
- PERIA: Perceive, Reason, Imagine, Act via Holistic Language and Vision Planning for Manipulation13
Comprehensive evaluations across three task domains demonstrate that PERIA, benefiting from holistic language and vision planning, significantly outperforms competitive baselines in both instruction following accuracy and task success rate on complex manipulation tasks.
- GraSS: Combining Graph Neural Networks with Expert Knowledge for SAT Solver Selection13
GraSS is presented, a novel approach for automatic SAT solver selection based on tripartite graph representations of instances and a heterogeneous graph neural network (GNN) model, which enrich the graph representation with domain-specific decisions, such as novel node feature design, positional encodings for clauses in the graph, and a GNN architecture tailored to the authors' tripartite graphs.
- Lightweight Neural App Control13
A novel mobile phone control architecture, Lightweight Multi-modal App Control (LiMAC), for efficient interactions and control across various Android apps by integrating a small Action Transformer integrated with a fine-tuned vision-language model (VLM) for real-time decision-making and task execution.
- 13
- Pessimistic value iteration for multi-task data sharing in Offline Reinforcement Learning13
An uncertainty-based MTDS approach that shares the entire dataset without data selection is proposed, which provides a unified framework for single- and multi-task offline RL and shows the optimality gap of the method is only related to the expected data coverage of the shared dataset, thus resolving the distribution shift issue in data sharing.
- 13
- 13
- Efficient Multiagent Policy Optimization Based on Weighted Estimators in Stochastic Cooperative Environments13
WDDQN outperforms an existing DRL algorithm (double DQN) and an MA-DRL algorithm (lenient Q-learning) regarding the averaged reward and the convergence speed and is more likely to converge to the Pareto-optimal Nash equilibrium in stochastic cooperative environments.
- 13
- 13
- Heuristic Collective Learning for Efficient and Robust Emergence of Social Norms13
Experimental results show that both learning strategies can support the emergence of desirable social norms more efficiently in a much broader range of multiagent interaction scenarios than previous work, and also are robust across different network topologies.
- Hierarchical task network-enhanced multi-agent reinforcement learning: Toward efficient cooperative strategies12
Hierarchical Symbolic Multi-Agent Reinforcement Learning (HS-MARL) is introduced, a novel approach that incorporates hierarchical knowledge into MARL to effectively reduce the exploration space and significantly outperforms other methods in environments with sparse rewards and suboptimal conditions.
- Iteratively Refined Behavior Regularization for Offline Reinforcement Learning12
This paper proposes a new algorithm that substantially enhances behavior-regularization based on conservative policy iteration, and proves that in the tabular setting this algorithm is capable of learning the optimal policy covered by the offline dataset.
- 12
- 12
- A Deep Learning-based Genome-wide Polygenic Risk Score for Common Diseases Identifies Individuals with Risk12
DeepPRS performs better than the other two existing state-of-art methods on Alzheimer's disease, inflammatory bowel disease, type 2 diabetes and breast cancer and has the chance to identify high-risk individuals even with few known risk SNPs.
- Efficient Deep Reinforcement Learning through Policy Transfer12
A novel Policy Transfer Framework (PTF) is proposed that learns when and which source policy is the best to reuse for the target policy and when to terminate it by modeling multi-policy transfer as the option learning problem.
- Bayes-ToMoP: A Fast Detection and Best Response Algorithm Towards Sophisticated Opponents12
Experimental results show both Bayes-ToMoP and deep Baye's ToMoP outperform the state-of-the-art approaches when faced with different types of opponents in two-agent competitive games.
- Interactions in Multiagent Systems: Fairness, Social Optimality and Individual Rationality12
The book describes fundamental and advanced techniques of how multi-agent systems can be engineered towards the goal of ensuring fairness, social optimality, and individual rationality.
- Designing minimal effective normative systems with the help of lightweight formal methods12
This work proposes a framework that facilitates designing of minimal effective normative systems using lightweight formal methods and manually establishes an induction proof so as to lift the design to an arbitrary number of agents.
- Introducing decision entrustment mechanism into repeated bilateral agent interactions to achieve social optimality12
A novel learning strategy TaFSO is proposed which combines the characteristics of both teacher and follower strategies to effectively influence the opponent’s behavior towards SOSNE outcomes by exploiting their limitations.
- Probabilistic Model Checking Multi-agent Behaviors in Dispersion Games Using Counter Abstraction12
This paper model the system using discrete-time Markov chain (DTMC) and reduce the state space of the models by applying counter abstraction technique and demonstrates that probabilistic model checking technique is applicable, and indeed useful for automatic analysis and verification of multi-agent dynamics.
- Embodied Arena: A Comprehensive, Unified, and Evolving Evaluation Platform for Embodied AI11
Embodied Arena is presented, a comprehensive, unified, and evolving evaluation platform for Embodied AI that establishes a systematic embodied capability taxonomy spanning three levels (perception, reasoning, task execution), seven core capabilities, and 25 fine-grained dimensions, enabling unified evaluation with systematic research objectives.
- Unified and explainable molecular representation learning for imperfectly annotated data from the hypergraph view11
OmniMol is presented, a framework using hypergraphs to improve predictions of molecular properties, addressing challenges of imperfect data annotation and enhancing model explainability, and achieves state-of-the-art performance in properties prediction.
- PCBAgent: An Agent-based Framework for High-Density Printed Circuit Board Placement11
This work introduces a novel agent-based framework that automatically generates PCB layouts meeting industrial constraints through user interactions, and shows that this framework outperforms other state-of-the-art methods.
- MENTOR: Guiding Hierarchical Reinforcement Learning With Human Feedback and Dynamic Distance Constraint11
This work proposes a general hierarchical reinforcement learning framework incorporating human feedback and dynamic distance constraints, termed MENTOR, which acts as a “mentor” and proposes the Dynamic Distance Constraint (DDC) mechanism dynamically adjusting the space of optional subgoals, such that MENTOR can generate subgoals matching the low-level policy learning process from easy to hard.
- Enhancing Robotic Manipulation with AI Feedback from Multimodal Large Language Models11
This study trains a multimodal LLM, termed CriticGPT, capable of understanding trajectory videos in robot manipulation tasks, serving as a critic to offer analysis and preference feedback, and validate the effectiveness of preference labels generated by CriticGPT from a reward modeling perspective.
- 11
- 11
- 11
- 11
- 11
- 11
- Enhancing Logical Reasoning in Large Language Models through Graph-based Synthetic Data10
It is demonstrated that supervised fine-tuning (SFT) with synthetic graph-based reasoning data effectively enhances LLMs' reasoning performance without compromising their effectiveness on other standard evaluation benchmarks.
- 10
- RITA: Boost Driving Simulators with Realistic Interactive Traffic Flow10
This work proposes Realistic Interactive TrAffic flow (RITA) as an integrated component of existing driving simulators to provide high-quality traffic flow for the evaluation and optimization of the tested driving strategies, and demonstrates its capacity to create diversified and high-fidelity traffic simulations in several highly interactive highway scenarios.
- Transformer in Transformer as Backbone for Deep Reinforcement Learning10
The Transformer in Transformer (TIT) backbone is proposed, which cascades two Transformers in a very natural way: the inner one is used to process a single observation, while the outer one is responsible for processing the observation history; combining both is expected to extract spatial-temporal representations for good decision-making.
- 10
- Breaking the Curse of Dimensionality in Multiagent State Space: A Unified Agent Permutation Framework10
This work proposes two novel designs to achieve permutation invariant (PI), one of which applies a hypernetwork to generate customized embedding for each component, which has higher representational capacity than the previous embedding-sharing method.
- 10
- Mastering Basketball With Deep Reinforcement Learning: An Integrated Curriculum Training Approach10
An integrated curriculum training approach (ICTA) for enhancing the cooperation ability of the basketball team is proposed, which is a curriculum switcher that focuses on learning the switch of the cooperative curriculum within one team by taking over collaborative actions such as passing from a single-player's action spaces.
- 10
- 10
- Machine Learning Insides OptVerse AI Solver: Design Principles and Applications9
This study presents a comprehensive study on the integration of machine learning techniques into Huawei Cloud's OptVerse AI Solver, which aims to mitigate the scarcity of real-world mathematical programming instances, and to surpass the capabilities of traditional optimization techniques.
- MFE-ETP: A Comprehensive Evaluation Benchmark for Multi-modal Foundation Models on Embodied Task Planning9
This work attempts to provide an in-depth and comprehensive evaluation of the performance of MFM s on embodied task planning, aiming to shed light on their capabilities and limitations in this domain.
- 9
- 9
- AhaRobot: A Low-Cost Open-Source Bimanual Mobile Manipulator for Embodied AI8
Despite its low cost, AhaRobot enables imitation learning of complex household behaviors involving bimanual coordination, upper-body mobility, and contact-rich interaction, with data quality comparable to VR-based collection.
- Scalable In-Context Q-Learning8
A prompt-based multi-head transformer architecture that simultaneously predicts optimal policies and in-context value functions using separate heads using separate heads is designed, enabling the construction of a compact prompt that facilitates fast and precise in-context inference.
- A Transfer Approach Using Graph Neural Networks in Deep Reinforcement Learning8
This paper proposes TURRET (Transfer Using gRaph neuRal nETworks), to utilize the generalization capabilities of Graph Neural Networks (GNNs) to facilitate efficient and effective multi-source policy transfer learning in the state-action mismatch setting.
- 8
- 8
- Dual-Process Graph Neural Network for Diversified Recommendation8
The Dual-Process Graph Neural Network (DPGNN) is proposed, an efficient diversity-enhanced recommender system, resonating with the dual-process model of human cognition and the arousal theory of human interest.
- 8
- 8
- 8
- 8
- Defending Against Man-In-The-Middle Attack in Repeated Games8
This work proposes practical adaptive algorithms for the defenders and the attacker to learn towards the unique Nash equilibrium through repeated interactions and shows that the algorithms are able to converge to Nash equilibrium strategy efficiently.
- 8
- 8
- An Adaptive Learning Framework for Efficient Emergence of Social Norms: (Extended Abstract)8
Experimental results indicate that the proposed general learning framework outperforms the static learning framework in various comparison criteria.
- 7
- MODULI: Unlocking Preference Generalization via Diffusion Models for Offline Multi-Objective Reinforcement Learning7
MODULI (Multi-objective Diffusion Planner with Sliding Guidance), which employs a preference-conditioned diffusion model as a planner to generate trajectories that align with various preferences and derive action for decision-making, and introduces two return normalization methods under diverse preferences for refining guidance.
- Multiagent Gumbel MuZero: Efficient Planning in Combinatorial Action Spaces7
The new algorithms, MA Gumbel AlphaZero and MA Gumbel MuZero, respectively without and with model learning, achieve superior performance on cooperative multiagent control problems, while reducing the number of environmental interactions by up to an order of magnitude compared to model-free approaches.
- 7
- 7
- 7
- Transfer Reinforcement Learning Based Negotiating Agent Framework7
This work presents a novel Transfer Learning based Negotiating Agent (TLNAgent) framework that allows a negotiating agent to transfer previous knowledge from source strategies optimized by deep reinforcement learning, to boost its performance in new tasks.
- ASN: action semantics network for multiagent reinforcement learning7
A novel network architecture, named Action Semantics Network (ASN), that explicitly represents different actions’ influence on other agents using neural networks based on the action semantics between them and can be easily combined with existing deep reinforcement learning (DRL) algorithms to boost their performance.
- 7
- 7
- 7
- 7
- Critic PI2: Master Continuous Planning via Policy Improvement with Path Integrals and Deep Actor-Critic Reinforcement Learning7
A novel model-based reinforcement learning frameworks called Critic PI2 is presented, which combines the benefits from trajectory optimization, deep actor-critic learning, and model- based reinforcement learning to achieve a new state of the art in a range of challenging continuous domains.
- Recurrent Deep Multiagent Q-Learning for Autonomous Agents in Future Smart Grid7
This paper develops an effective pricing strategy for brokers in local electricity retail market based on recurrent deep multiagent reinforcement learning and sequential clustering and demonstrates the effectiveness of the reward shaping mechanism.
- Adaptive Defending Strategy for Smart Grid Attacks7
The preliminary results show that the amount of load shedding costs can be significantly reduced by employing an AMS over a NE strategy, which is theoretically guaranteed to converge to a best response strategy against any stationary attacker.
- Learning to Achieve Socially Optimal Solutions in General-Sum Games7
A novel learning strategy TaFSO is proposed which can effectively influence the opponent's behavior towards socially optimal outcomes by utilizing the characteristic of best-response learners.
- From Chaos to Order: The Atomic Reasoner Framework for Fine-grained Reasoning in Large Language Models6
Experimental results demonstrate AR's superior reasoning capabilities without the computational burden of exhaustive solution searches, particularly excelling in linguistic logic puzzles, and substantiate AR's effectiveness in enhancing LLMs' capacity for robust, long-sequence logical reasoning and deliberation.
- More than A Point: Capturing Uncertainty with Adaptive Affordance Heatmaps for Spatial Grounding in Robotic Tasks6
RoboMAP is proposed, a framework that represents spatial targets as continuous, adaptive affordance heatmaps that captures the uncertainty in spatial grounding and provides richer information for downstream policies, thereby significantly enhancing task success and interpretability.
- 6
- DR-Label: Improving GNN Models for Catalysis Systems by Label Deconstruction and Reconstruction6
A novel graph neural network (GNN) supervision and prediction strategy DR-Label is presented, which enhances the supervision signal, reduces the multiplicity of solutions in edge representation, and encourages the model to provide node predictions that are graph structural variation robust.
- 6
- A novel framework for single-cell Hi-C clustering based on graph-convolution-based imputation and two-phase-based feature extraction6
A novel framework, named ScHiC-Rep, for singlecell Hi-C data representation and clustering is proposed, which outperforms existing state-of-the-art approaches on both human and mouse datasets.
- 6
- Large-Scale Home Energy Management Using Entropy-Based Collective Multiagent Reinforcement Learning Framework6
An Entropy-Based Collective Multiagent Deep Reinforcement Learning (EB-C-MADRL) framework is presented and experiments demonstrate that EB-C-MADRL can reduce both the long-term group power consumption cost and daily peak demand effectively compared with existing approaches.
- Building Personalized Simulator for Interactive Search6
A simulator is proposed to employ to mimic the environment for the offline training of the agent and modelbased approach is used to train the simulator and is able to use the data efficiently.
- Learning to Advertise with Adaptive Exposure via Constrained Two-Level Reinforcement Learning.6
This paper investigates the problem of advertising with adaptive exposure, in which the number of ad slots and their locations can dynamically change over time based on their relative scores with recommendation products, and proposes a constrained two-level reinforcement learning to decouple the original advertising exposure optimization problem into two relatively independent sub-optimization problems.
- Fairness in Cooperative Multiagent Systems6
To achieve effective coordinations among agents, one traditional approach is to employ a superagent to determine the behaviors for all other agents in the system, but there exist a number of disadvantages by adopting this approach.
- Achieving Social Optimality with Influencer Agents6
Simulation results show that a small proportion of influencer agents can significantly increase the average percentage of socially optimal outcomes attained in the system and better performance can be achieved compared with that of previous work.
- Incorporating Fairness into Agent Interactions Modeled as Two-Player Normal-Form Games6
A fairness model is proposed which incorporates two important aspects of fairness motivations, and the solution concept of fairness equilibrium is defined, and it is shown that the predictions of the model successfully reflect the intuitions from both aspects of unfair motivations.
- Trajectory World Models for Heterogeneous Environments5
This work explores pre-training world models for heterogeneous environments by addressing key transfer barriers in both data diversity and model flexibility, and introduces UniTraj, a unified dataset comprising over one million trajectories from 80 environments, designed to scale data while preserving critical diversity.
- Improving Generalization in Offline Reinforcement Learning via Latent Distribution Representation Learning5
This paper introduces the Latent Distribution Representation Learning (LAD) framework, which aims to characterize the multiple latent distributions within offline data and reduce the distribution gaps between any pair of them, and derive a generalization error bound to support LAD theoretically and verify its effectiveness through extensive experiments.
- 5
- JigsawPlanner: Jigsaw-like Floorplanner for Eliminating Whitespace and Overlap among Complex Rectilinear Modules5
JigsawPlanner is introduced, a novel and flexible Jigsaw-like floorPlanner that facilitates floorplanning to handle complex-shaped rectilinear modules and can effectively eliminate whitespace and overlaps simultaneously and significantly reduce the Half Perimeter Wire Length compared to state-of-the-art baselines.
- GFlowNets with Human Feedback5
The goal of GFlowHF is to learn a policy that is strictly proportional to human ratings, instead of only focusing on human favorite ratings like RLHF, and experiments show that G flowHF can achieve better exploration ability than RLHF.
- 5
- Accelerating deep reinforcement learning via knowledge-guided policy network5
A knowledge-guided policy network, a novel framework that combines suboptimal human knowledge with reinforcement learning and RL that significantly improves the learning efficiency of basic RL algorithms, even with very low-performance human prior knowledge.
- 5
- 5
- Foresee then Evaluate: Decomposing Value Estimation with Latent Future Prediction5
This paper analytically decomposes the value function into a latent future dynamics part and a policy-independent trajectory return part, inducing a way to model latent dynamics and returns separately in value estimation.
- 5
- An Adaptive Learning Based Network Selection Approach for 5G Dynamic Environments5
An adaptive learning based strategy is proposed, which enables users to adaptively adjust their selections in response to the gradually or abruptly changing environment, and is experimentally shown to converge to Nash equilibrium, which also turns out to be both Pareto optimal and socially optimal.
- 5
- Socially-Aware Multiagent Learning: Towards Socially Optimal Outcomes5
A novel gradient ascent based algorithm (SA-IGA) which augments the basic gradient-ascent algorithm by incorporating social awareness into the policy update process, and results show that an SA-PGA agent can achieve higher social welfare and also is robust against individually rational opponents by reaching Nash equilibrium solutions.
- 5
- 5
- 5
- Unveiling the Entropy Dynamics of Chain-of-Thought Reasoning4
This paper investigates the entropy dynamics of Chain-of-Thought (CoT) and uncovers a consistent two-phase structure: an Uncertainty Region of exploration transitioning sharply to a Confidence Region of convergence, and develops a training-free framework for real-time inference control.
- Few-Shot Vision-Language Action-Incremental Policy Learning4
TOPIC pioneers few-shot continual learning in the robotic manipulation task, and extensive experimental results demonstrate that TOPIC outperforms state-of-the-art baselines by over 26$\%$ in success rate, significantly enhancing the continual learning capabilities of existing Transformer-based policies.
- 4
- The MMO Economist: AI Empowers Robust, Healthy, and Sustainable P2W MMO Economies4
This paper introduces a novel hierarchical Reinforcement Learning (RL) solution for achieving Pareto optimality between profitability and equality in P2W economies and establishes an economic simulation environment that facilitates authentic and realistic assessments of MMO economic evolution.
- A Hierarchical Imitation Learning-based Decision Framework for Autonomous Driving4
Through an evaluation in several typical driving scenarios within the SMARTS autonomous driving simulation environment, the proposed hierarchical decision-making framework significantly outperforms end-to-end reinforcement learning algorithms and behavior cloning algorithm, achieving an average pass rate of over 90%.
- T3S: Improving Multi-Task Reinforcement Learning with Task-Specific Feature Selector and Scheduler4
A novel MTRL framework called Task-Specific feature Selector and Scheduler (T3S) is proposed, which consists of two components: a feature selector and a task scheduler that consistently outperforms the state-of-the-art M TRL algorithms on various robotics manipulation tasks.
- 4
- 4
- Limited Information Opponent Modeling4
Experimental results demonstrate that LIOM is able to accurately extract opponent policy representations even when the opponent’s information is limited, and has a certain degree of generalization ability for unknown policies, outperforming existing opponent modeling algorithms.
- SCC-rFMQ: a multiagent reinforcement learning method in cooperative Markov games with continuous actions4
An sample based independent learning method, named Sample Continuous Coordination with recursive Frequency Maximum Q-Value (SCC-rFMQ), which divides the multiagent cooperative problem with continuous actions into two layers and can handle cooperative problems in continuous action cooperative Markov games effectively.
- 4
- Low-Bias Amplification for Robust DNA Data Readout4
The study clearly demonstrated that canonical polymerase chain reaction created significant DNA amplification biases, which greatly hinder fast and stable data retrieving from hundred-thousand synthetic DNA sequences encoding over 2.85 megabyte (MB) digital data.
- Cross-data Automatic Feature Engineering via Meta-learning and Reinforcement Learning4
This work presents a framework called Cross-data Automatic Feature Engineering Machine (CAFEM), which formalizes the FE problem as an optimization problem over a Feature Transformation Graph (FTG) and shows that FeL outperforms existing approaches and is robust on the selection of learning algorithms.
- Reinforcement Learning for Cooperative Overtaking4
This paper proposes two basic coordination models to enable distributed learning of cooperative overtaking maneuvers in a group of vehicles and proves that, by capturing the underlying consistency of identities or positions during vehicles' movement, efficient coordinated behaviors can be achieved simply through vehicles' local learning interactions.
- Reinforcement Social Learning of Coordination in Networked Cooperative Multiagent Systems4
This work investigates the multiagent coordination problems in cooperative environments under the networked social learning framework focusing on two representative topologies: the small-world and the scale-free network.
- Reinforcement Social Learning of Coordination in Cooperative Multiagent Systems (Extended Abstract)4
This paper investigates the multiagent coordination problems in cooperative environments under the social learning framework, in which there exists a large population of agents and each agent interacts with another agent randomly in each round.
- An Efficient Negotiation Protocol to Achieve Socially Optimal Allocation4
An efficient protocol within a more general negotiation framework, which can lead agents to achieve socially optimal allocation and theoretically prove that the final allocation is guaranteed to be socially optimal provided that the agents are altruistic-individually rational.
- Towards A Unified Policy Abstraction Theory and Representation Learning Approach in Markov Decision Processes3
This work proposes a unified policy abstraction theory, containing three types of policy abstraction and explaining their partial ordering relationship, and proposes a policy representation learning approach and policy optimization algorithm based on deep metric learning.
- 3
- Exploiting counter-examples for active learning with partial labels3
This paper builds an intuitive baseline that can be seamlessly incorporated into existing AL frameworks, and proposes a simple but effective WorseNet to directly learn from this complementary pattern, leveraging the distribution gap between WorseNet and the predictor to enhance both the performance of the predictor itself and the sample selection process.
- Protein Sequence Design by Entropy-based Iterative Refinement3
An iterative sequence refinement pipeline, which can refine the sequence generated by existing sequence design models, and selects and retains reliable predictions based on the model’s confidence in predicted distributions, and decodes the residue type based on a partially visible environment.
- 3
- Prototypical context-aware dynamics generalization for high-dimensional model-based reinforcement learning3
A Prototypical Context-Aware Dynamics (ProtoCAD) model is proposed, which captures the local dynamics by time consistent latent context and enables dynamics generalization in high-dimensional control tasks.
- 3
- 3
- MGHRL: Meta Goal-Generation for Hierarchical Reinforcement Learning3
This paper proposes a new meta-RL algorithm called Meta Goal-generation for Hierarchical RL (MGHRL), which learns to generate high-level meta strategies over subgoals given past experience and leaves the rest of how to achieve subGoals as independent RL subtasks.
- 3
- 3
- SCC-rFMQ Learning in Cooperative Markov Games with Continuous Actions3
An independent learner hierarchical method, named Sample Continuous Coordination with recursive Frequency Maximum Q-Value (SCC-rFMQ), which divides the coordination problem into two layers and can handle coordination problems in continuous action cooperative Markov games effectively.
- The Dynamics of Opinion Evolution in Gossiper-Media Model with WoLS-CALA Learning3
This work proposes a reinforcement learning framework to model the interactions between the public ( aka gossipers) and the media agents and theoretically proves that WoLS-CALA can learn to Nash equilibria for two-agent games with continuous action space.
- Automated Software Security Requirements Recommendation Based on FT-SR Model3
Experimental results show that the proposed approach can generate a set of recommended SRs reducing theulty of SRs recommending even for people lack knowledge of security.
- Negotiating with Unknown Opponents Toward Multi-lateral Agreement in Real-Time Domains3
A novel negotiation approach is proposed that enables an agent to obtain efficient agreements in this challenging multi-lateral negotiations under real-time constraints by employing sparse pseudo-input Gaussian processes to model opponents, and adopting an adaptive decision-making mechanism to handle uncertainty in negotiation.
- 3
- 2
- Evolving LLMs' Self-Refinement Capability via Synergistic Training-Inference Optimization2
This paper introduces a novel post-training and inference framework, called ARIES, which iteratively performs preference training and self-refinement-based data collection and surpasses powerful models such as GPT-4o under the self-refinement setting.
- 2
- SeaDAG: Semi-autoregressive Diffusion for Conditional Directed Acyclic Graph Generation2
SeaDAG is introduced, a semi-autoregressive diffusion model for conditional generation of Directed Acyclic Graphs (DAGs), and explicitly train the model to learn graph conditioning with a condition loss, which enhances the diffusion model's capacity to generate graphs that are both realistic and aligned with specified properties.
- 2
- The Ladder in Chaos: A Simple and Effective Improvement to General DRL Algorithms by Policy Path Trimming and Boosting2
This paper studies how the policy networks of typical DRL agents evolve during the learning process by empirically investigating several kinds of temporal change for each policy parameter, and proposes a simple and effective method, called Policy Path Trimming and Boosting (PPTB), as a general plug-in improvement to DRL algorithms.
- Hierarchical Task Network Planning for Facilitating Cooperative Multi-Agent Reinforcement Learning2
SOMARL is presented, a framework that uses prior knowledge to reduce the exploration space and assist learning in sparse reward multi-agent reinforcement learning environments with traps and reports superior performance over state-of-the-art MARL and subgoal-based baselines for MARL environments significantly.
- Multi-agent Policy Reciprocity with Theoretical Guarantee2
A novel multi-agent policy reciprocity (PR) framework is proposed, where each agent can fully exploit cross-agent policies even in mismatched states, and theoretical analysis shows that agents can asymptotically reach consensus through individual perceived rewards and converge to an optimal value function, which implies the stability and effectiveness of PR.
- Uncertainty-Aware Low-Rank Q-Matrix Estimation for Deep Reinforcement Learning2
A novel Uncertainty-Aware Low-rank Q-matrix Estimation (UA-LQE) algorithm is proposed as a general framework to facilitate the learning of value function and reveals a positive correlation between value matrix rank and value estimation uncertainty.
- Decomposed Mutual Information Optimization for Generalized Context in Meta-Reinforcement Learning2
The theoretical analysis shows that DOMINO can overcome the underestimation of the mutual information caused by multi-confounded challenges via learning disentangled context and reduce the demand for the number of samples collected in various environments.
- 2
- 2
- 2
- Represent Your Own Policies: Reinforcement Learning with Policy-extended Value Function Approximator2
- ONECG: Online Negotiation Environment for Coalitional Games2
This demo presents an online negotiation environment for coalitional games (ONECG), in which coalitional negotiation can be conduced in a distributed way between people, agents, or in mixed settings via offer exchange and natural language communication.
- Explicitly Coordinated Policy Iteration2
This work proposes the explicitly coordinated policy iteration (EXCEL) algorithm which always forces agents to coordinate by comparing the agents’ separated optimistic and average value functions and proposes three solutions for deep reinforcement learning extensions of EXCEL.
- Efficient meta reinforcement learning via meta goal generation2
This work proposes a new meta-RL algorithm called meta goal-generation for hierarchical RL (MGHRL) by leveraging hierarchical actor-critic framework, which enables more efficient and effective meta-learning from past experience and outperforms state-of-the-art meta- RL and Hierarchical-RL methods in sparse reward settings.
- 2
- 2
- Circuit Synthesis based on Hierarchical Conditional Diffusion1
A layer-wise autoregressive diffusion model that leverages circuit hierarchy information, a differentiable Function Decoder that helps ensure the functional correctness of the circuits, and a Monte Carlo Tree Search-based optimization algorithm for post-generation refinement are introduced.
- 1
- A Theory of Multi-Agent Generative Flow Networks1
The theory framework of MA-GFlowNets, which can be applied to multiple agents to generate objects collaboratively through a series of joint actions, is proposed and experimental results demonstrate the superiority of the proposed framework compared to reinforcement learning and MCMC-based methods.
- Advancing Symbolic Discovery on Unsupervised Data: A Pre-training Framework for Non-degenerate Implicit Equation Discovery1
A novel pre-training framework -- namely, Pre-trained neural symbolic model for Implicit Equation (PIE) -- is introduced to discover implicit equations from unsupervised data and shows an encouraging step towards general scientific discovery on unsupervised data.
- 1
- 1
- 1
- 1
- 1
- 1
- On the Convergence Theory of Meta Reinforcement Learning with Personalized Policies1
This paper proposes a novel personalized Meta-RL (pMeta-RL) algorithm, which aggregates task-specific personalized policies to update a meta-policy used for all tasks, while maintaining customized policies to maximize the average return of each task under the constraint of the meta- policy.
- 1
- 1
- 1
- 1
- 1
- Learning to Accelerate Heuristic Searching for Large-Scale Maximum Weighted b-Matching Problems in Online Advertising1
This work designs a multichannel graph neural network to predict the threshold of the matched edges, by which the search region could be significantly reduced, and proposes a parallel heuristic search algorithm to iteratively improve the solution quality until convergence.
- An Optimal Rewiring Strategy for Cooperative Multiagent Social Learning1
This paper proposes an optimal rewiring strategy to select most beneficial peers to maximize the accumulated payoffs in long-run interactions and empirically demonstrates the effects of the approach in a variety of large-scale MASs.
- 1
- 1
- Effective norm emergence in cell systems under limited communication1
This work proposes a learning approach which can trade off the agents’ performance of coordinating on a consistent norm and the communication cost involved and significantly reduces the coordination dependence among agents.
- Achieving Multiagent Coordination Through CALA-rFMQ Learning in Continuous Action Space1
A novel algorithm called CALA-rFMQ (Continuous Action Learning Automata with recursive Frequency Maximum Q-Value) is proposed that ensures robust and efficient coordination among multiple agents in continuous action spaces and outperforms previous works.
- 1
- An Improved Android Collusion Attack Detection Method Based on Program Slicing1
An Android collusion attack sensitive information leakage detection tools called CollusionDetector, which can obtain better results than existing tools and methods and can more accurately find the path of privacy propagation.
- Fepchecker: An Automatic Model Checker for Verifying Fairness and Non-Repudiation of Security Protocols in Web Service1
In the model checking algorithm, the fairness and non-repudiation properties are verified based on Labeled Transition System (LTS) semantics model and the MakeOneMove method is used to explore the state space on-the-fly in the verification process.
- Reinforcement social learning of social optimality with influencer agents1
Simulation results show that a small proportion of influencer agents can significantly increase the average percentage of socially optimal outcomes attained in the system and better performance can be achieved compared with that of previous work.
- 1
- 1
- 1
- PACE: Unleashing the Power of Code Embeddings to Boost AutoML Agents–
PACE (Pre-execution Admission via Code Embeddings), an online-adaptive admission control framework that improves budgeted sample efficiency by estimating candidate utility prior to execution from within-run execution history, without training a separate offline predictor.
- –
- –
- –
- –
- EVLP:Learning Unified Embodied Vision-Language Planner with Reinforced Supervised Fine-Tuning–
EVLP (Embodied Vision-Language Planner) is presented, an innovative multimodal unified generation framework that jointly models linguistic reasoning and visual generation that achieves multimodal planning for long-horizon tasks through a novel training pipeline incorporating dynamic pretraining and reinforced alignment.
- –
- –
- Generalizable and Relation Sensitive Netlist Representation for Analog Circuit Design–
A novel state representation for analog circuits is proposed to capture the heterogeneity of connections between different components to reduce runtime significantly in both fine-tuning and zero-shot transfer settings compared to current SOTA baselines.
- –
- –
- –
- Reweighted Interacting Langevin Diffusions: an Accelerated Sampling Methodfor Optimization–
A new technique to accelerate sampling methods for solving difficult optimization problems with Langevin dynamics, and an interacting particle scheme that approximates a Reweighted Interacting Langevin Diffusion system (RILD) is proposed.
- –
- –
- –
- State-Aware Proximal Pessimistic Algorithms for Offline Reinforcement Learning–
A principled algorithmic framework for offline RL called SA-PP, which is leveraging discounted stationary state distribution ratios between the learning policy and the offline dataset to modulate the degree of behavior regularization in a state-wise manner, so that pessimism can be implemented in a more appropriate way.
- –
- Opponent modeling with trajectory representation clustering–
This paper extracts the representation of opponent policies and makes explicit clustering distinctions through the contrastive learning autoencoder and demonstrates the effectiveness of the method under a classical opponent modeling environment (soccer) and shows the clustering effect of different opponent policies.
- Offline-to-Online Co-Evolutional User Simulator and Dialogue System–
An Offline-to-oNline Co-Evolutional (ONCE) framework is introduced, which enables bias-aware concurrent joint update for RL-based fine-tuning whilst takes advantages from GPT-2 based end- to-end modeling on US and DS.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- A Prediction and Learning Based Approach to Network Selection in Dynamic Environments–
This paper proposes a prediction and learning based approach, which considers both the fluctuation of radio resource and the variation of user demand, and achieves significantly better performance compared with two existing approaches in terms of load balancing, user payoff and the overall bandwidth utilization efficiency.
- Reciprocal Social Strategy in Social Repeated Games and Emergence of Social Norms–
A new repeated game mechanism for modeling multi-agent interactions as well as a new reinforcement learning based agent learning method are proposed, which differs from traditional repeated games in that the agents are anonymous, and the agents interact with randomly chosen opponents during each iteration.
- –
- –
- –
- –
- Background and Previous Work–
Previous work in multiagent learning literature is explored by dividing them into two major categories: cooperativemultiagent systems and competitive multiagent systems, distinguished by the underlying intentional stance of the agents within the system.
- –
- Social Optimality in Competitive Multiagent Systems–
This chapter looks at how to handle the question of how socially optimal solutions can be achieved in competitive MASs which consist of individually rational agents within two major learning frameworks in MASs.
- Individual Rationality in Competitive Multiagent Systems–
This chapter focuses on the first research direction by considering an important competitive multiagent interaction scenario: bilateral negotiation, and the second research direction will be the focus of Chap.
- Spatial evolutionary game-theoretic perspective on agent-based complex negotiations–
It is shown how spatial evolutionary game theory can be used to interpret bilateral negotiation results among state-of-the-art strategies in a new class of negotiation games that take negotiation locality into consideration.
- Robustness Analysis of Negotiation Strategies through Multiagent Learning in Repeated Negotiation Games–
The repeated negotiation game learning framework is proposed to evaluate the robustness of different negotiation strategies, in which each agent can adopt any rational learning approach to make decisions without knowing the global payoff information beforehand.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.