Weinan Zhang

Manipulation as in Simulation: Enabling Accurate Geometry Perception in Robots

Your Agent May Misevolve: Emergent Risks in Self-evolving LLM Agents

Audio Turing Test: Benchmarking the Human-likeness of Large Language Model-based Text-to-Speech Systems in Chinese

Iterative Refinement of Flow Policies in Probability Space for Online Reinforcement Learning

CATArena: Evaluation of LLM Agents Through Iterative Tournament Competitions

USE: Enhancing Mixed-Motive Cooperation via Unified Self and Collective Rewards

CoreCodeBench: A Configurable Multi-Scenario Repository-Level Benchmark

InfoDeepSeek: Benchmarking Agentic Information Seeking for Retrieval-Augmented Generation

MARFT: Multi-Agent Reinforcement Fine-Tuning

LoopTool: Closing the Data–Training Loop for Robust LLM Tool Calls

RAD: Retrieval High-quality Demonstrations to Enhance Decision-making

Unified Latent Steering and Residual Refinement for Online Improvement of Diffusion Policy Models

ATGen: Adversarial Reinforcement Learning for Test Case Generation

Progra: Progress-Aware Reinforcement Learning for Multi-Turn Function Calling

CoLD: Counterfactually-Guided Length Debiasing for Process Reward Models in Mathematical Reasoning

Improve LLM Pre-training with RL-Guided Annealing

APTBench: Benchmarking Agentic Potential of Base LLMs During Pre-Training

TRACEBench: Personalized Function Calling Benchmark Based on Real-World Human Interaction

Information-Theoretic Reward Decomposition for Generalizable RLHF

KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills

GenPO: Generative Diffusion Models Meet On-Policy Reinforcement Learning

Uni-RL: Unifying Online and Offline RL via Implicit Value Regularization

Robust Function-Calling for On-Device Language Model via Function Masking

Flexible Realignment of Language Models

AgentNet: Decentralized Evolutionary Coordination for LLM-based Multi-Agent Systems

ReMA: Learning to Meta-Think for LLMs with Multi-agent Reinforcement Learning

Reconstruction-Guided Policy: Enhancing Decision-Making through Agent-Wise State Consistency

Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport

MobileUse: A Hierarchical Reflection-Driven GUI Agent for Autonomous Mobile Operation

ContraDiff: Planning Towards High Return States via Contrastive Learning

Large Language Models are Demonstration Pre-Selectors for Themselves

DyDiff: Long-Horizon Rollout via Dynamics Diffusion for Offline Reinforcement Learning

Large Language Models are Demonstration Pre-Selectors for Themselves

RethinkMCTS: Refining Erroneous Thoughts in Monte Carlo Tree Search for Code Generation

Computing Ex Ante Equilibrium in Heterogeneous Zero-Sum Team Games