AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

What Does Vision Tool-Use Reinforcement Learning Really Learn? Disentangling Tool-Induced and Intrinsic Effects for Crop-and-Zoom

Yan Ma, Weiyu Zhang, Tianle Li, Du Linge, Xuyang Shen, Pengfei Liu

ICML 2026 · Reinforcement learning › Visual and representation RL · approach: Robust visual encoders for RL · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
show the tool schema but forbid execution (vs plain tool-free prompt)Qwen2.5-VL-7B · Average of 6 benchmarksaccuracy: 48.4 -> 42.6 (-5.8)Hurt
show the tool schema but forbid execution (vs plain tool-free prompt)Qwen3-VL-8B · Average of 6 benchmarksaccuracy: 53.0 -> 40.0 (-13.0)Hurt
show the tool schema but forbid execution (vs plain tool-free prompt)Qwen3-VL-8B · VStaraccuracy: 82.7 -> 57.1Hurt
allow tool execution (vs tool-free inference)Qwen2.5-VL-7B · Average of 6 benchmarksaccuracy: 48.4 -> 42.2Hurt
allow tool execution (vs tool-free inference)Qwen3-VL-8B · Average of 6 benchmarksaccuracy: 53.0 -> 61.2Improved
allow tool execution (vs tool-free inference)Qwen2.5-VL-7B · VisualProbe Easyaccuracy: 39.0 -> 21.3Hurt
allow tool execution (vs tool-free inference)Qwen3-VL-8B · VisualProbe Easyaccuracy: 41.8 -> 56.7Improved
train with tools removed during RL (no-tool RL vs tool-RL), tool-free accuracy (ablation)Qwen2.5-VL-7B · final Acc_woaccuracy: 0.644 vs 0.636No clear effect
train with tools removed during RL (no-tool RL vs tool-RL), tool-free accuracy (ablation)Qwen3-VL-8B · final Acc_woaccuracy: 0.678 vs 0.681No clear effect
add +0.1 tool-call bonus to the reward (ablation)Qwen2.5-VL-7B · MED termsCall Gain and Call Harm: +0.1 bonus for correct responses that call the toolMixed
remove tool access in RL training (ablation)Qwen2.5-VL-7B · MED termsCall Gain and Call Harm: both call-side terms close to zeroHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy