paper-with-me

홈 › Papers

Stabilizing On-Policy Distillation for MLLM Reasoning with Global Normalization

2026-06-08 · Dongze Hao, Zhiwei Jin, Chen Chen, Haonan Lu arxiv

On-policy distillation (OPD) has recently emerged as an important post-training paradigm. By using a stronger teacher model to provide dense, fine-grained supervision for sampled trajectories, OPD offers a clear advantage over reinforcement learning with verifiable rewards (RLVR), which typically depends on sparse binary or outcome-based environmental feedback. However, naive token-level distillation can suffer from gradient instability, due to magnitude misalignment in outlier states. To address this issue, we propose Globally Normalized Distillation Policy Optimization (GNDPO), a practical method that stabilizes optimization by transforming raw KL scores into batch-level relative advantages. This normalization effectively mitigates gradient explosions while retaining the benefits of token-level guidance. Experimental results show that GNDPO substantially improves training robustness and downstream performance across multimodal reasoning tasks. The code is released at https://github.com/OPPO-Mente-Lab/GNDPO.

📄 PDF Abstract BibTeX arXiv:2606.09091

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement LearningMultimodal Reasoning

Similar Papers 제목 키워드 기반

OPD-V: Visual On-Policy Self-Distillation with Modality Balance

2026-08-05 · Aniri, Jinhe Bi, Peng Liao, Zengjie Jin 외 hf

On-Policy Self-Distillation (OPSD) has become a standard post-training approach for improving visual reasoning in multimodal large language models (MLLMs). Existing methods draw privileged information from diverse input …

Visual Reasoning

Seeing Before Reasoning: Decoupling Perception and Reasoning for Shortcut-Resilient Multimodal On-Policy Self-Distillation

2026-06-17 · Sihan Wang, Xiyao Liu, Lianqing Liu, Zhi Han arxiv

On-policy self-distillation (OPSD) trains a model on its own rollouts and uses a frozen copy to provide dense token-level targets conditioned on a reference target. This works well for LLM reasoning, but a direct extensi…

Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation

2026-05-18 · Qianhao Yuan, Jie Lou, Xing Yu, Hongyu Lin 외 arxiv

Multimodal Large Language Models (MLLMs) still struggle with fine-grained visual understanding, where answers often depend on small but decisive evidence in the full image. We observe a regional-to-global perception gap:…

IPIBench: Evaluating Interactive Proactive Intelligence of MLLMs under Continuous Streams

2026-05-26 · Jinzhao Li, Yinuo Chen, Wenxuan Song, Yijia Lei 외 arxiv

Recent multimodal large language models (MLLMs) achieve strong performance on reactive question answering, but real-world streaming assistants require proactive reasoning over continuous visual inputs. Existing benchmark…

Question Answering

Constitutional On-Policy Safe Distillation

2026-06-02 · Ming Wen, Yuxuan Liu, Kun Yang, Yunhao Feng 외 arxiv

On-policy self-distillation (OPSD) has emerged as an efficient post-training paradigm by using a teacher conditioned on privileged information to provide dense token-level supervision. Prior work has shown that OPSD can …