AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

ATP-Bench: Towards the Agentic Tool Planning for MLLM Interleaved Generation

Yinuo Liu, Zi Qian, Heng Zhou, Jiahao Zhang, Yajie Zhang, Zhihang Li, Mengyu Zhou, Erchao Zhao, xiaoxi jiang, Guanjun Jiang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Multimodal LLM benchmarks

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
add 3 in-context tool demonstrations (ablation)GPT-4o · ATP-Bench 800-query subsetFinal Score: 60.35 -> 73.19Improved
add 3 in-context tool demonstrations (ablation)GPT-4o · ATP-Bench 800-query subsetSuccess Rate: 63.05 -> 82.01Improved
add 3 in-context tool demonstrations (ablation)Qwen2.5-VL-72B · ATP-Bench 800-query subsetSuccess Rate: 39.06 -> 72.88Improved
add 3 in-context tool demonstrations (ablation)Qwen2.5-VL-72B · ATP-Bench 800-query subsetFinal Score: 53.88 -> 72.86Improved
add 3 in-context tool demonstrations (ablation)LLaMA-3.2-11B · ATP-Bench 800-query subsetFinal Score: 29.60 -> 30.20No clear effect
change judge model (Claude Sonnet 4.5, Gemini 2.5 Pro, GPT-5) (ablation)MAM · ATP-Bench subsetmodel rankingNo clear effect
Search Enhanced vs Diffusion Enhanced prompt boundary (ablation)Claude Sonnet 4.5, Grok-4.1, Gemini 3 Pro · ATP-Bench subsettool adoption rateMixed
model size/type: Gemini 3 Pro vs other MLLMsGemini 3 Pro · ATP-BenchFinal Score: 69.34 -> 79.88Improved
MAM judge agreement with human evaluationMAM · end-to-end human evalSpearman rho (FS): rho=0.8909Improved
compare tool-set F1 ranking with MAM Final Scoreground-truth F1 · ATP-BenchSpearman rho: rho=0.879Improved

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