Zhiheng Xi

AgentGym-RL: An Open-Source Framework to Train LLM Agents for Long-Horizon Decision Making via Multi-Turn RL

Game-RL: Synthesizing Multimodal Verifiable Game Data to Boost VLMs' General Reasoning

BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping

Critique-RL: Training Language Models For Critiquing Through Two-Stage Reinforcement Learning

Better know nothing than half-know anything: A Precise and Efficient Dataset for Scientific Reasoning in Language Models

Why Reinforcement Fine-Tuning Enables MLLMs Preserve Prior Knowledge Better: A Data Perspective

Two Minds Better Than One: Collaborative Reward Modeling for LLM Alignment

From Scores to Preferences: Redefining MOS Benchmarking for Speech Quality Reward Modeling

Pre-Trained Policy Discriminators are General Reward Models

Have the VLMs Lost Confidence? A Study of Sycophancy in VLMs

Distill Visual Chart Reasoning Ability from LLMs to MLLMs

RMB: Comprehensively benchmarking reward models in LLM alignment

AgentGym: Evaluating and Evolving Large Language Model-based Agents across Diverse Envronments

Progressive LLM Alignments Using Two-Player Games