Diversity Over Frequency: Rethinking Tool Use in Visual Chain-of-Thought Agents

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Dong-Hee, Tan, Reuben, Kim, Donghyun
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916068046405632
author Kim, Dong-Hee
Tan, Reuben
Kim, Donghyun
author_facet Kim, Dong-Hee
Tan, Reuben
Kim, Donghyun
contents Visual agents employ external visual tools within visual chains of thought to incorporate fine-grained evidence. While prior work has mainly studied these tools in visual search tasks, their role in more complex visual reasoning remains underexplored. In this paper, we move beyond simple visual search tasks to investigate more challenging tasks, including 3D spatial reasoning and medical visual question answering, where agents must integrate tool-acquired local evidence with the global context. We identify a {tool-use collapse phenomenon: models progressively stop using tools while still achieving higher task accuracy. Moreover, we observe a clear asymmetry: (i) completely eliminating tool use degrades performance, whereas (ii) incentivizing tool use yields only marginal gains despite substantially increasing usage. We find that vanilla training and tool-use encouragement both reduce rollout diversity, explaining why higher tool use does not yield stronger reasoning performance. Motivated by these findings, we add an entropy regularization term to encourage diverse rollout exploration, achieving the best performance despite gradually declining tool usage. % We further observe similar dynamics on medical VQA, suggesting that tool-use collapse is not limited to 3D spatial reasoning. Overall, our findings suggest a training-time view of tools as scaffolding, where broader exploration over language generation and visual tool invocation improves reasoning despite tool-use collapse. Project page: https://scaffolded-exploration.github.io
format Preprint
id arxiv_https___arxiv_org_abs_2606_00096
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Diversity Over Frequency: Rethinking Tool Use in Visual Chain-of-Thought Agents
Kim, Dong-Hee
Tan, Reuben
Kim, Donghyun
Computer Vision and Pattern Recognition
Artificial Intelligence
Visual agents employ external visual tools within visual chains of thought to incorporate fine-grained evidence. While prior work has mainly studied these tools in visual search tasks, their role in more complex visual reasoning remains underexplored. In this paper, we move beyond simple visual search tasks to investigate more challenging tasks, including 3D spatial reasoning and medical visual question answering, where agents must integrate tool-acquired local evidence with the global context. We identify a {tool-use collapse phenomenon: models progressively stop using tools while still achieving higher task accuracy. Moreover, we observe a clear asymmetry: (i) completely eliminating tool use degrades performance, whereas (ii) incentivizing tool use yields only marginal gains despite substantially increasing usage. We find that vanilla training and tool-use encouragement both reduce rollout diversity, explaining why higher tool use does not yield stronger reasoning performance. Motivated by these findings, we add an entropy regularization term to encourage diverse rollout exploration, achieving the best performance despite gradually declining tool usage. % We further observe similar dynamics on medical VQA, suggesting that tool-use collapse is not limited to 3D spatial reasoning. Overall, our findings suggest a training-time view of tools as scaffolding, where broader exploration over language generation and visual tool invocation improves reasoning despite tool-use collapse. Project page: https://scaffolded-exploration.github.io
title Diversity Over Frequency: Rethinking Tool Use in Visual Chain-of-Thought Agents
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2606.00096