QIANG FU

One-Shot Style Personalization for RL Agents via Latent Discriminator

DANCE: Difficulty and Novelty Co-driven Exploration

MCU: An Evaluation Framework for Open-Ended Game Agents

Diverse Policies Recovering via Pointwise Mutual Information Weighted Imitation Learning

Goal-Oriented Skill Abstraction for Offline Multi-Task Reinforcement Learning

Online-to-Offline RL for Agent Alignment

Optimizing Latent Goal by Learning from Trajectory Preference

Offline Opponent Modeling with Truncated Q-driven Instant Policy Refinement

Enhance Reasoning for Large Language Models with Reinforcement Learning in the Game Werewolf

Towards Evaluating Generalist Agents: An Automated Benchmark in Open World

Training Open-ended Policies to follow Video-prompt Instructions with Reinforcement Learning