Research map / Image generation / Text-to-image diffusion
Paper2Figure: A Multi-Agent Collaborative System for Figure Generation Towards Academic Research Paper
Siwei Han, Haonian Ji, Siyang Xin, Juanquan Shi, Shi Qiu, Xinyu Ye, Peng Xia, Jiaqi Liu, Zhaorun Chen, Yiyang Zhou, Linjie Li, Lijuan Wang and 1 more
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 |
|---|---|---|---|
| add Refinement Agents on top of Generation Agents (ablation) | Paper2Figure · Paper2Figure Bench | Avg. score: 74.8 -> 79.2 | Improved |
| add Refinement Agents (ablation) | Paper2Figure · Paper2Figure Bench | Completeness (C): 71.0 -> 77.5 | Improved |
| add Refinement Agents (ablation) | Paper2Figure · Paper2Figure Bench | A2 relation consistency: 69.0 -> 74.5 | Improved |
| add Refinement Agents (ablation) | Paper2Figure · Paper2Figure Bench | B1 layout and grouping: 83.0 -> 82.5 | No clear effect |
| generate figures in FigScript with multi-agents instead of direct SVG from the strongest LLM | GPT-4o agents · Paper2Figure Bench | Avg. score: 65.1 -> 79.2 | Improved |
| use Mermaid instead of SVG output for LLM figure generation | GPT-5 · Paper2Figure Bench | Avg. score: 65.1 -> 51.3 | Hurt |
| use text-to-image model instead of code-based generation | GPT-Image-1 · Paper2Figure Bench | Avg. score: 65.1 -> 34.9 | Hurt |
| use a more capable LLM for SVG generation (GPT-5 vs GPT-4o) | SVG baselines · Paper2Figure Bench | Avg. score: 59.2 -> 65.1 | Improved |
| use the rubric-based LLM metric instead of BERTScore for agreement with humans | Paper2Figure Bench metric · Human agreement | Pearson: 0.6337 -> 0.7345 | Improved |
| use Gemini 2.5 Flash instead of Gemini 2.5 Pro for SVG generation | SVG baselines · Paper2Figure Bench | Avg. score: 60.5 -> 55.1 | Hurt |
Closest papers by meaning
- Paper2Poster: Towards Multimodal Poster Automation from Scientific Papers (NeurIPS 2025 · similarity 0.89)
- P2P: Automated Paper-to-Poster Generation and Fine-Grained Benchmark (ICLR 2026 · similarity 0.89)
- From Words to Structured Visuals: A Benchmark and Framework for Text-to-Diagram Generation and Editing (CVPR 2025 · similarity 0.86)
- Presenting a Paper is an Art: Self-Improvement Aesthetic Agents for Academic Presentations (ICLR 2026 · similarity 0.85)
- Muses: 3D-Controllable Image Generation via Multi-Modal Agent Collaboration (AAAI 2025 · similarity 0.84)
- GenArtist: Multimodal LLM as an Agent for Unified Image Generation and Editing (NeurIPS 2024 · similarity 0.84)
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.
