AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

VisNumBench: Evaluating Number Sense of Multimodal Large Language Models

Tengjin Weng, Jingyi Wang, Wenhao Jiang, Zhong Ming

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 15 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
fine-tune InternVL2-8B with MMPR preference data (MPO) (ablation)InternVL2-8B · VisNumBench-Syntheticaccuracy (%): 39.56 -> 40.65Improved
fine-tune InternVL2-8B with MMPR preference data (MPO) (ablation)InternVL2-8B · VisNumBench-Realaccuracy (%): 39.58 -> 39.91No clear effect
fine-tune LLaVA-v1.5-13B on MathV360K (Math-LLaVA) (ablation)LLaVA-v1.5-13B · VisNumBench-Syntheticaccuracy (%): 32.15 -> 35.81Improved
fine-tune LLaVA-v1.5-13B on MathV360K (Math-LLaVA) (ablation)LLaVA-v1.5-13B · VisNumBench-Realaccuracy (%): 40.02 -> 33.15Hurt
add chain-of-thought training (LLaVA-o1-100k) (ablation)Llama-VL-3.2-11B · VisNumBench-Syntheticaccuracy (%): 43.92 -> 45.50Improved
add chain-of-thought training (LLaVA-o1-100k) (ablation)Llama-VL-3.2-11B · VisNumBench-Realaccuracy (%): 43.24 -> 38.36Hurt
add R1-style rule-based RL reasoning training (R1-Onevision) (ablation)Qwen2.5-VL-7B · VisNumBench-Syntheticaccuracy (%): 46.19 -> 38.87Hurt
add R1-style rule-based RL reasoning training (R1-Onevision) (ablation)Qwen2.5-VL-7B · VisNumBench-Realaccuracy (%): 41.02 -> 38.25Hurt
add CoT prompt (ablation)Gemini 2.0 Flash · VisNumBench-Syntheticaccuracy (%): 57.57 -> 60.14Improved
add CoT prompt (ablation)Qwen2.5-VL-72B · VisNumBench-Syntheticaccuracy (%): 58.46 -> 56.68Hurt
add CoT prompt (ablation)InternVL2.5-78B · VisNumBench-Syntheticaccuracy (%): 56.18 -> 53.21Hurt
add CoT prompt (ablation)Qwen2.5-VL-72B · VisNumBench-Realaccuracy (%): 53.33 -> 51.11Hurt

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