5 papers
SteerRM: Debiasing Reward Models via Sparse Autoencoders
Mengyuan Sun, Zhuohao Yu, Weizheng Gu +2
Reward models (RMs) are critical components of alignment pipelines, yet they exhibit biases toward superficial stylistic cues, preferring better-presented responses over semantical…
What Do Agents Learn from Trajectory-SFT: Semantics or Interfaces?
Weizheng Gu, Chengze Li, Zhuohao Yu +6
Large language models are increasingly evaluated as interactive agents, yet standard agent benchmarks conflate two qualitatively distinct sources of success: semantic tool-use and…
SAEMark: Steering Personalized Multilingual LLM Watermarks with Sparse Autoencoders
Zhuohao Yu, Xingru Jiang, Weizheng Gu +4
Watermarking LLM-generated text is critical for content attribution and misinformation prevention. However, existing methods compromise text quality, require white-box model access…
RewardAnything: Generalizable Principle-Following Reward Models
Zhuohao Yu, Jiali Zeng, Weizheng Gu +7
Reward Models, essential for guiding Large Language Model optimization, are typically trained on fixed preference datasets, resulting in rigid alignment to single, implicit prefere…
Reasoning Through Execution: Unifying Process and Outcome Rewards for Code Generation
Zhuohao Yu, Weizheng Gu, Yidong Wang +5
Large Language Models excel at code generation yet struggle with complex programming tasks that demand sophisticated reasoning. To bridge this gap, traditional process supervision…