arXiv cs.AI· Junxin Wang, Yingfeng Luo, Xiangchen Song, Weiqiao Shan, Runsong Zhao, Xinyu Liu, Shunjie Xing, Yifu Huo, Dingyang Lin, Qiuyu Gu, Yongyu Mu, Chenglong Wang, Changliang Li, Jingbo Zhu, Tong Xiao·· 12 小时前AI 评分29
PEAR：面向大型视觉语言模型感知与探索的自适应优势重塑方法
PEAR: Adaptive Advantage Reshaping for Perception and Exploration in Large Vision-Language Models
AI 导读
PEAR 是一种面向多模态推理的自适应优势重塑方法，以视觉依赖度为基础信号，仅在高熵位置叠加熵驱动的探索增益，视觉依赖度越低增益越大，融合信号再映射为 token 级正权重以重塑序列级优势。在 Qwen2.5-VL-3B/7B 上，PEAR 在多模态推理基准上的平均表现优于所评估的基线，并在文本推理和通用多模态基准上取得领先结果。
正文 · 原文
Authors:Junxin Wang, Yingfeng Luo, Xiangchen Song, Weiqiao Shan, Runsong Zhao, Xinyu Liu, Shunjie Xing, Yifu Huo, Dingyang Lin, Qiuyu Gu, Yongyu Mu, Chenglong Wang, Changliang Li, Jingbo Zhu, Tong Xiao
Abstract:Effective multimodal reasoning requires a model to reliably utilize visual information while sufficiently exploring diverse reasoning paths. To achieve this goal, prior work combines visual dependency and token entropy to reshape sequence-level advantages, but scaling the contribution of token entropy by visual dependency can leave high-entropy tokens with low visual dependency underweighted, despite their important role in reasoning exploration. To address this limitation, we propose PEAR, an adaptive advantage reshaping method for improving perception and exploration in multimodal reasoning. PEAR retains visual dependency as the base signal and adds an entropy-based exploration gain only at high-entropy positions, with lower visual dependency leading to a larger additional gain. The fused signal is then mapped to positive token-level weights that reshape the sequence-level advantage. Experiments on Qwen2.5-VL-3B/7B show that PEAR achieves the highest average performance among evaluated baselines across multimodal reasoning benchmarks, alongside leading results on textual reasoning and general multimodal benchmarks.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.32216 [cs.AI] |
| (or arXiv:2610.32216v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.32216 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Junxin Wang [view email]
[v1]
Fri, 23 Oct 2026 06:14:15 UTC (2,902 KB)
来源：arXiv cs.AI · arxiv.org