Liangyou Li

From Verifiable Dot to Reward Chain: Harnessing Verifiable Reference-based Rewards for Reinforcement Learning of Open-ended Generation

ToolACE-MT: Non-Autoregressive Generation for Agentic Multi-Turn Interaction

Bridging and Modeling Correlations in Pairwise Data for Direct Preference Optimization

RevisEval: Improving LLM-as-a-Judge via Response-Adapted References

Subtle Errors Matter: Preference Learning via Error-injected Self-editing