Research map / Vision-language and multimodal / Visual question answering
Diagram-MMU: A Multi-Modal Benchmark for Scientific Diagrams
Weihao Bo, Shan Zhang, Yanpeng Sun, Jie Liu, Yongke Yao, Jinhao Du, Wei He, KAI ZOU, Zechao Li, Jingdong Wang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| provide diagram objects as context for editing (ablation) | all six agentic models · Diagram-MMU D2C-E mini | F1avg preserve: gains 4.1-10.6 | Improved |
| provide diagram objects as context for editing (ablation) | all six agentic models · Diagram-MMU D2C-E mini | F1avg edit: gains 3.3-7.1 | Improved |
| provide diagram objects as context for parsing (ablation) | Gemini-3.0 Pro, Qwen3-VL-8B · Diagram-MMU D2C-P mini | F1avg: -2.0 and -6.1 | Hurt |
| provide diagram objects as context for parsing (ablation) | Claude-4.6 Opus, Gemini-3.1 Pro · Diagram-MMU D2C-P mini | CrystalBLEU: +0.3 and +1.4 | Improved |
| provide diagram objects as context for question answering (ablation) | GPT-5.2 · Diagram-MMU DQA mini | Accuracy: -7.3 | Hurt |
| add TikZ search tool for parsing (ablation) | Gemini-3.1 Pro · Diagram-MMU D2C-P mini | F1avg: -4.3 | Hurt |
| add TikZ search tool (ablation) | Claude-4.6 Opus · Diagram-MMU D2C-P mini | F1avg: +2.2 | Improved |
| generate TikZ code first, then edit (ablation) | Claude-4.6 Opus, GPT-5.2 · Diagram-MMU D2C-E mini | F1avg preserve/edit: +2.5/+2.6 and +2.5/+3.5 | Improved |
| generate TikZ code first, then answer (ablation) | Claude-4.6 Opus · Diagram-MMU DQA mini | Accuracy: +3.8 | Improved |
| generate TikZ code first, then answer (ablation) | other models · Diagram-MMU DQA mini | Accuracy: -1.3 to -3.2 | Hurt |
| plan tool use with objects for parsing (ablation) | Claude-4.6 Opus · Diagram-MMU D2C-P mini | F1avg/CBLEU: +5.1/+2.8 | Improved |
| plan tool use with objects for parsing (ablation) | Gemini-3.1 Pro, Qwen3-VL-8B · Diagram-MMU D2C-P mini | F1avg/CBLEU: -5.0/-4.5 and -5.6/-6.4 | Hurt |
Closest papers by meaning
- CoG-DQA: Chain-of-Guiding Learning with Large Language Models for Diagram Question Answering (CVPR 2024 · similarity 0.88)
- MME-SCI: A Comprehensive and Challenging Science Benchmark for Multimodal Large Language Models (AAAI 2026 · similarity 0.88)
- From Words to Structured Visuals: A Benchmark and Framework for Text-to-Diagram Generation and Editing (CVPR 2025 · similarity 0.86)
- DiGraphHal-Bench: Evaluating Multimodal Large Language Models on Complex Directed Graphs (CVPR 2026 · similarity 0.86)
- ChartMuseum: Testing Visual Reasoning Capabilities of Large Vision-Language Models (NeurIPS 2025 · similarity 0.85)
- VCG-Bench: Towards A Unified Visual-Centric Benchmark for Structured Generation and Editing (ICML 2026 · similarity 0.85)
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.
