VC-Tooler：学习组合式与自适应视觉工具使用
VC-Tooler: Learning Compositional and Adaptive Visual Tool Use
AI总结：
VC-Tooler是一种学习组合式与自适应视觉工具使用的模型，通过分层合成轨迹库与两阶段训练，在通用和具身基准上达到开源模型最优性能，且具良好迁移能力。
AI中文摘要：
具身多模态推理通过让视觉语言模型（VLMs）借助视觉工具交互主动获取和细化视觉证据，拓展了被动图像理解。有效的视觉工具使用需要三项能力：将工具调用基于视觉语境、跨多步骤组合工具、根据工具返回的观测结果调整推理。然而现有方法大多聚焦于固定工具空间内的基础调用模式，对组合性与适应性的处理不足。本文提出VC-Tooler，将视觉工具使用作为组合式与自适应能力进行学习。为此，我们首先通过分层合成管线构建轨迹库，涵盖单工具基础、多工具组合、多样工具语境与接口三个能力层级；随后分两阶段训练模型：建立上述能力的监督冷启动阶段，以及鼓励准确、高效、语境感知的视觉工具使用的强化学习阶段。VC-Tooler在通用与具身基准的开源模型中达到最优性能，包括V*上的95.8%和VTC-Bench上的35.3%，且在推理时更丰富的工具设置下展现出良好的迁移能力。项目页面：this https URL
英文摘要：
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 $95.8\%$ on V* and $35.3\%$ on VTC-Bench, and shows promising transfer under richer tool settings at inference time. Project page: https://w1zheng.github.io/VC-Tooler
