arXiv cs.AI· Tengda Guo, Jie Leng, Hanlei Li, Yaoyuan Liang, Qingyue Zhang, Dian Yang, Mingyu Zhang, Yuhua Fu, Shao-Lun Huang·· 13 小时前AI 评分29
CAVE：面向碎片化视觉证据推理的结构化信用分配方法
CAVE: A Structured Credit Assignment Approach for Fragmented Visual Evidence Reasoning
AI 导读
研究者提出 CAVE，一种基于 GRPO 的结构化过程奖励方法，通过信念更新、证据获取和自适应焦点控制三类信号在动作层面评估中间步骤贡献，用于交错式视觉推理。团队同时构建 TRACER-Bench，覆盖四个非局部且语义易混淆的推理维度并提供关键中间证据。实验显示 CAVE 在碎片化视觉证据整合任务上显著提升，在通用多模态评测上保持竞争力。
正文 · 原文
Abstract:Vision-Language Models (VLMs) have achieved strong performance on general multimodal reasoning, yet remain challenged in integrating nonlocal visual information to support semantically underdetermined visual reasoning. We describe this challenge as Fragmented Visual Reasoning. To this end, we propose Credit Assignment for Visual Evidence (CAVE), a structured process-reward method based on GRPO for interleaved visual reasoning. Specifically, CAVE evaluates the contribution of intermediate steps at the action level via three complementary reasoning process signals: belief update, evidence acquisition, and adaptive focus control, thereby guiding the model to optimize each reasoning action and learn more reliable visual reasoning strategies. Meanwhile, we construct TRACER-Bench, which covers four nonlocal and semantically confusable reasoning dimensions and provides key intermediate evidence to supervise reasoning paths. Experiments demonstrate that CAVE substantially improves performance on tasks requiring fragmented visual evidence integration, covering both public benchmarks and our newly introduced TRACER-Bench, while retaining competitive performance on general multimodal evaluations. Further analyses reveal that CAVE effectively improves the visual reasoning capacity and exhibits stronger robustness under longer-range and deeper cross-region dependencies.
| Comments: | 24 pages, 6 figures. Preprint |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.16416 [cs.CV] |
| (or arXiv:2605.16416v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.16416 arXiv-issued DOI via DataCite |
Submission history
From: Tengda Guo [view email]
[v1]
Wed, 13 May 2026 16:50:24 UTC (22,726 KB)
[v2]
Wed, 21 Oct 2026 09:09:33 UTC (22,726 KB)
来源：arXiv cs.AI · arxiv.org