# ToolVision: Learning When and How to Use Visual Tools with Capability-Aligned Supervision

arXiv: https://arxiv.org/abs/2608.08907 · HTML: https://mlsanity.com/paper/2608.08907/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 24 of its 50 references that are in the map.
- First posted on arXiv 9 Aug 2026.

## Builds on (most cited first, 24 of 24)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning](https://mlsanity.com/paper/2203.10244/) (2022)
- [Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning](https://mlsanity.com/paper/2503.09516/) (2025)
- [DocVQA: A Dataset for VQA on Document Images](https://mlsanity.com/paper/2007.00398/) (2020)
- [A Diagram Is Worth A Dozen Images](https://mlsanity.com/paper/1603.07396/) (2016)
- [Visual-RFT: Visual Reinforcement Fine-Tuning](https://mlsanity.com/paper/2503.01785/) (2025)
- [V*: Guided Visual Search as a Core Mechanism in Multimodal LLMs](https://mlsanity.com/paper/2312.14135/) (2023)
- [ReTool: Reinforcement Learning for Strategic Tool Use in LLMs](https://mlsanity.com/paper/2504.11536/) (2025)
- [ToolRL: Reward is All Tool Learning Needs](https://mlsanity.com/paper/2504.13958/) (2025)
- [MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?](https://mlsanity.com/paper/2408.13257/) (2024)
- [Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models](https://mlsanity.com/paper/2406.09403/) (2024)
- [Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontiers](https://mlsanity.com/paper/2506.23918/) (2025)
- [OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning](https://mlsanity.com/paper/2501.00321/) (2024)
- [Learning To Count Everything](https://mlsanity.com/paper/2104.08391/) (2021)
- [GRIT: Teaching MLLMs to Think with Images](https://mlsanity.com/paper/2505.15879/) (2025)
- [Grounded Reinforcement Learning for Visual Reasoning](https://mlsanity.com/paper/2505.23678/) (2025)
- [PyVision: Agentic Vision with Dynamic Tooling](https://mlsanity.com/paper/2507.07998/) (2025)
- [CodeDance: A Dynamic Tool-integrated MLLM for Executable Visual Reasoning](https://mlsanity.com/paper/2512.17312/) (2025)
- [Thinking with Programming Vision: Towards a Unified View for Thinking with Images](https://mlsanity.com/paper/2512.03746/) (2025)

## Built on by (most cited first, 1 of 1)

- [Aphanta: Diagnosing Task-Aligned Image-Edited Intermediates for Multimodal Reasoning](https://mlsanity.com/paper/2608.26993/) (2026)

## Abstract (arXiv)

Thinking with images allows a multimodal model to compensate for limited perception by invoking visual tools through code. Yet the prevailing SFT-then-RL recipe creates a different supervision misalignment at each stage. SFT is expected to teach how to use tools, but trajectories from stronger teachers may succeed through perceptual capabilities that a smaller student cannot reliably reproduce or exploit, causing the student to imitate tool-call patterns without learning how to make them useful. RL is expected to teach when to use tools, but outcome-only rewards make fallible tool execution a liability and suppress tool use, whereas a blanket bonus for every correct tool-using trajectory encourages valid but ineffective operations. To address these two misalignments, we introduce ToolVision. During SFT, a multi-agent pipeline explores candidate trajectories, and a committee including student-scale models scores stepwise evidence gain to rank and prune the search branches. Only successfully executed trajectories with correct final answers are retained for SFT. Before RL, ToolVision compares the learner's performance with and without tools, then rewards successful tool use only on questions where tools provide a clear benefit. Both signals are constructed automatically from public task data without additional human annotations of tool use or necessity. ToolVision-8B improves over its base on all seven main benchmarks, surpasses Thyme-7B, CodeVision-8B, and CodeDance-7B on all three high-resolution benchmarks, and outperforms Qwen3-VL-32B-Thinking on V* and HRBench 8K. We will publicly release the datasets and source code.