Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

CAVE: A Structured Credit Assignment Approach for Fragmented Visual Evidence Reasoning

About

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.

Tengda Guo, Jie Leng, Hanlei Li, Yaoyuan Liang, Qingyue Zhang, Dian Yang, Mingyu Zhang, Yuhua Fu, Shao-Lun Huang• 2026

Related benchmarks

TaskDatasetResultRank
Mathematical ReasoningMathVerse mini
Accuracy55.7
83
Multimodal ReasoningRealworldQA
Mean@8 Accuracy69.7
40
Mathematical ReasoningMathVista mini
Score71.2
37
Multimodal ReasoningBLINK
Accuracy61.5
20
Multimodal ReasoningMMStar
Score65.2
18
Vision ReasoningTunnel Vision
ObjReID Score53.5
15
Vision ReasoningBabyVision
FD Score15.3
15
Vision ReasoningTunnel Vision, BabyVision, and TRACER-Bench
Overall Score33.3
15
Vision ReasoningTRACER-Bench
LT Score31.4
15
Hallucination EvaluationHallusionBench Image
Average Score64.1
5
Showing 10 of 10 rows

Other info

Follow for update