7 papers
Prefix Teach, Suffix Fade: Local Teachability Collapse in Strong-to-Weak On-Policy Distillation
Kaiyuan Liu, Ziyuan Zhuang, Yang Bai +3
On-policy distillation (OPD) trains a student model on its own rollouts using dense feedback from a stronger teacher. Prior literature suggests that, provided teacher feedback is a…
Multi-Objective and Mixed-Reward Reinforcement Learning via Reward-Decorrelated Policy Optimization
Yang Bai, Kaiyuan Liu, Ziyuan Zhuang +5
Complex reinforcement learning environments frequently employ multi-task and mixed-reward formulations. In these settings, heterogeneous reward distributions and correlated reward…
Unveiling Fine-Grained Visual Traces: Evaluating Multimodal Interleaved Reasoning Chains in Multimodal STEM Tasks
Jing Jin, Hao Liu, Yan Bai +9
Multimodal large language models (MLLMs) have shown promising reasoning abilities, yet evaluating their performance in specialized domains remains challenging. STEM reasoning is a…
A Survey on LLM Mid-Training
Chengying Tu, Xuemiao Zhang, Rongxiang Weng +6
Recent advances in foundation models have highlighted the significant benefits of multi-stage training, with a particular emphasis on the emergence of mid-training as a vital stage…
AdaR: A Framework for Equipping LLMs with Adaptive Reasoning
Zhejian Lai, Xiang Geng, Zhijun Wang +7
Mathematical reasoning is a primary indicator of large language models (LLMs) intelligence. However, existing LLMs exhibit failures in robustness and generalization. This paper att…
Libra: Assessing and Improving Reward Model by Learning to Think
Meng Zhou, Bei Li, Jiahao Liu +5
Reinforcement learning (RL) has significantly improved the reasoning ability of large language models. However, current reward models underperform in challenging reasoning scenario…
Length Desensitization in Direct Preference Optimization
Wei Liu, Yang Bai, Chengcheng Han +5
Direct Preference Optimization (DPO) is widely utilized in the Reinforcement Learning from Human Feedback (RLHF) phase to align Large Language Models (LLMs) with human preferences,…