AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

MMReason: An Open-Ended Multi-Modal Multi-Step Reasoning Benchmark for MLLMs Toward AGI

Huanjin Yao, Jiaxing Huang, Yawen Qiu, Michael K. Chen, Wenzheng Liu, Wei Zhang, Wenjie Zeng, Xikun Zhang, Jingyi Zhang, YuXin Song, Wenhao Wu, Dacheng Tao

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 18 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
remove answer choices from the question (ablation)Qwen2-VL-7B · MathVistaaccuracy: 58.2 -> 37.8Hurt
remove answer choices from the question (ablation)GPT-4o · MathVistaaccuracy: 63.8 -> 40.8Hurt
apply multi-model voting filtering to candidate questions (ablation)GPT-4o · MMReason candidatestext-only accuracy: 13.4 -> 0.78Improved
apply multi-model voting filtering to candidate questions (ablation)GPT-4o · MMReason candidatesmultimodal relevance rate (%): 56.5 -> 97.0Improved
apply multi-model voting filtering to candidate questions (ablation)Qwen2.5-VL-7B · MMReason candidatesmultimodal relevance rate (%): 60.7 -> 96.1Improved
apply voting filtering, measured with text and visual input (ablation)GPT-4o · MMReason candidatesaccuracy: 30.8 -> 25.7Hurt
use newly collected questions instead of reformulated M3CoT questionsGPT-4o-1120 · MMReasonfinal answer reasoning accuracy: 28.9 -> 21.8Hurt
scale Qwen-2.5-VL from 7B to 72BQwen-2.5-VL · MMReasonfinal answer reasoning accuracy: 16.8 -> 24.7Improved
scale InternVL-2.5-MPO from 8B to 78BInternVL-2.5-MPO · MMReasonfinal answer reasoning accuracy: 12.4 -> 21.3Improved
use explicit chain-of-thought training (LLaVA-CoT, Mulberry) over base modelLLaMA-3.2-Vision-11B · MMReasonfinal answer reasoning accuracy: 11.9 -> 12.5No clear effect

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