# OPD-V：面向多模态大模型的视觉在策略自蒸馏新方法

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-08-05 08:00
- AIHOT 分数：50
- AIHOT 链接：https://aihot.virxact.com/items/cmsgv6bqi004vroxzftxcae7s
- 原文链接：https://arxiv.org/abs/2608.05131

## AI 摘要

OPD-V 提出一种视觉在策略自蒸馏（OPSD）范式，通过正负教师模型将模态平衡本身作为特权信息，并利用模态平衡对数间隔定义信任区域以筛选自蒸馏 token。在 6 个基准、4 种 MLLM 主干和 5 种后训练方法上的实验显示，OPD-V 在提升推理性能的同时降低了训练成本。

## 正文

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 sources to guide self-distillation. Yet these designs overlook Modality Imbalance, a challenge inherent to MLLM reasoning. When textual information dominates generation, the model cannot fully integrate its multimodal input. Consequently, carefully designed privileged information remains underused, limiting the effectiveness of OPSD. To examine this limitation, we construct a Positive Teacher with the Zoom-In Image and a Negative Teacher with the Mask Image, which exhibit different degrees of Modality Imbalance. Changes in their reasoning correctness and token logits reveal that Modality Balance can itself serve as privileged information. Motivated by this finding, we introduce OPD-V, a visual OPSD paradigm that instantiates such information through the Positive Teacher and Negative Teacher. Positive Modality-Balance Logits Margins define a Modality-Balance Trust Region that selects the on-policy tokens used for self-distillation. Experiments across 6 benchmarks, 4 MLLM backbones, and 5 post-training methods show that OPD-V consistently improves reasoning performance while reducing training cost.