SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.01334 · ICML · 2026

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

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

arXiv · PDF · Open in the Atlas

Code that ran

10 of 12 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 failures of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2602.01334.json · badge: /api/ran/2602.01334.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
GAIR-NLP/Med canonical (classifier label)
classifier basis: census
10 ran · 2 no recorded run
FunctionStatusWhere it lives
call_vision_tool_api Ran
ran; our contract draft was wrong, not the code
GAIR-NLP/Med/recipe/med/vision_tool_utils.py
pointer only (licence: NOASSERTION) · get_code("ef08eff3592eed93")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

compress_sequence Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
GAIR-NLP/Med/recipe/med/trajectory_saver.py
pointer only (licence: NOASSERTION) · get_code("7a6c9dc47a00f688")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

decode_base64_to_image Ran
ran, honoured the contract we drafted
GAIR-NLP/Med/recipe/med/vision_tool_utils.py
pointer only (licence: NOASSERTION) · get_code("c2f577aba1b58fee")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

encode_base64_image Ran
ran; our contract draft was wrong, not the code
GAIR-NLP/Med/recipe/med/vision_tool_utils.py
pointer only (licence: NOASSERTION) · get_code("6ba7eadbcd19027b")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

extract_and_compress_positions Ran
ran; our contract draft was wrong, not the code
GAIR-NLP/Med/recipe/med/trajectory_saver.py
pointer only (licence: NOASSERTION) · get_code("e5705fd63a13ad8e")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

extract_step_from_path Ran
ran, honoured the contract we drafted
GAIR-NLP/Med/recipe/med/analysis_plot/create_csv.py
pointer only (licence: NOASSERTION) · get_code("d18dae5bbc401ed2")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_baseline_model_from_exp_name Ran
ran; our contract draft was wrong, not the code
GAIR-NLP/Med/recipe/med/analysis_plot/create_csv.py
pointer only (licence: NOASSERTION) · get_code("801452f8375f0965")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

identify_model_from_exp_name Ran
ran; our contract draft was wrong, not the code
GAIR-NLP/Med/recipe/med/analysis_plot/calculate_and_plot_area.py
pointer only (licence: NOASSERTION) · get_code("a8ae2ec6f7be412f")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

load_experiment_data Ran
ran; our contract draft was wrong, not the code
GAIR-NLP/Med/recipe/med/analysis_plot/calculate_and_plot_area.py
pointer only (licence: NOASSERTION) · get_code("c83961fe47cdbec3")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

load_trajectory_jsonl Ran
ran; our contract draft was wrong, not the code
GAIR-NLP/Med/recipe/med/analysis_plot/create_csv.py
pointer only (licence: NOASSERTION) · get_code("950c26e69ae67ad9")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

extract_and_save_messages Not yet run
unverified (harvested, no recorded run)
GAIR-NLP/Med/recipe/med/trajectory_saver.py
pointer only (licence: NOASSERTION) · get_code("b5ee343e222cc933")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

load_baseline_performance Not yet run
unverified (harvested, no recorded run)
GAIR-NLP/Med/recipe/med/analysis_plot/calculate_and_plot_area.py
pointer only (licence: NOASSERTION) · get_code("5a917ff7861b94a2")
How it ran

The body of this sample is withheld: GitHub could not classify the licence of this paper's repository, GAIR-NLP/Med (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 12 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Vision tool-use reinforcement learning (RL) can equip vision-language models with visual operators such as crop-and-zoom and achieves strong performance gains, yet it remains unclear whether these gains are driven by improvements in tool use or evolving intrinsic capabilities. We introduce MED (Measure-Explain-Diagnose), a coarse-to-fine framework that disentangles intrinsic capability changes from tool-induced effects, decomposes the tool-induced performance difference into gain and harm terms, and probes the mechanisms driving their evolution. Across checkpoint-level analyses in the crop-and-zoom setting on two VLMs with different tool priors and six benchmarks, we find that improvements are dominated by intrinsic learning, while tooluse RL mainly reduces tool-induced harm (e.g., fewer call-induced errors and weaker tool schema interference) and yields limited progress in toolbased correction of intrinsic failures. Overall, in the crop-and-zoom setting studied here, current vision tool-use RL learns to coexist safely with tools rather than master them. Code: https: //github.com/GAIR-NLP/Med

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2602.01334")
get_code_for_paper("2602.01334")
have("2602.01334")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2602.01334.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2602.01334.svg)](https://syntology.ai/paper/2602.01334)

Connect an agent — have() is free.