English

VC-Tooler: Learning Compositional and Adaptive Visual Tool Use

Computer Vision and Pattern Recognition 2026-08-03 v1

Abstract

Agentic multimodal reasoning extends passive image understanding by allowing VLMs to actively acquire and refine visual evidence through visual tool interactions. Effective visual tool use requires three capabilities: grounding tool calls in visual context, composing tools across multiple steps, and adapting reasoning to tool-returned observations. However, existing approaches largely focus on grounding within fixed tool spaces and rigid invocation patterns, leaving composition and adaptation insufficiently addressed. We present VC-Tooler, which learns visual tool use as a compositional and adaptive capability. To this end, we first build a trajectory bank through a hierarchical synthesis pipeline covering three capability levels: single-tool grounding, multi-tool composition, and diverse tool contexts and interfaces. We then train the model in two stages: a supervised cold start that establishes these capabilities, followed by reinforcement learning that encourages accurate, efficient, and context-aware visual tool use. VC-Tooler achieves state-of-the-art performance among open-source models on both general-purpose and agentic benchmarks, including 95.8%95.8\% on V* and 35.3%35.3\% on VTC-Bench, and shows promising transfer under richer tool settings at inference time. Project page: https://w1zheng.github.io/VC-Tooler

Cite

@article{arxiv.2608.02217,
  title  = {VC-Tooler: Learning Compositional and Adaptive Visual Tool Use},
  author = {Yizheng Wu and Jiashen Hua and Bing Deng and Jieping Ye},
  journal= {arXiv preprint arXiv:2608.02217},
  year   = {2026}
}