Research map / Large language models / Agents, code and math
Automatic Method Illustration Generation for AI Scientific Papers via Drawing Middleware Creation, Evolution, and Orchestration
ZHUOLING Li, Jiarui Zhang, Ping Hu, Jason Kuen, Jiuxiang Gu, Hossein Rahmani, Jun Liu
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 drawing middleware repository (ablation) | FigAgent · FigAgentBench | VLM score: 39.5 -> 61.4 | Improved |
| add drawing middleware repository (ablation) | FigAgent · FigAgentBench | NA F1: 26.5 -> 51.4 | Improved |
| add Explore-and-Select strategy (ablation) | FigAgent · FigAgentBench | VLM score: 48.2 -> 61.4 | Improved |
| add Explore-and-Select instead of greedy selection (ablation) | FigAgent · FigAgentBench | VLM score: 54.3 -> 61.4 | Improved |
| add Explore-and-Select instead of greedy selection (ablation) | FigAgent · FigAgentBench | NA F1: 41.9 -> 51.4 | Improved |
| add middleware evolution (ablation) | FigAgent · FigAgentBench | VLM score: 45.5 -> 61.4 | Improved |
| add selection operation in evolution (ablation) | FigAgent · FigAgentBench | VLM score: 47.6 -> 61.4 | Improved |
| add mutation operation in evolution (ablation) | FigAgent · FigAgentBench | VLM score: 59.3 -> 61.4 | Improved |
| add crossover operation in evolution (ablation) | FigAgent · FigAgentBench | VLM score: 54.2 -> 61.4 | Improved |
| add new middlewares directly without evolution (ablation) | FigAgent (2023 repository) · FigAgentBench | VLM score: 49.5 -> 56.8 | Improved |
| add new middlewares via evolution (ablation) | FigAgent (2023 repository) · FigAgentBench | VLM score: 49.5 -> 64.9 | Improved |
| integrate new middlewares through evolution instead of direct adding (ablation) | FigAgent (2023 repository) · FigAgentBench | NA F1: 37.4 -> 50.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PaperBanana — PaperBanana: Automating Academic Illustration for AI Scientists (ICML 2026) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SketchAgent: Language-Driven Sequential Sketch Generation (CVPR 2025 · similarity 0.87)
- SketchMind: A Multi-Agent Cognitive Framework for Assessing Student-Drawn Scientific Sketches (NeurIPS 2025 · similarity 0.83)
- DialogDraw: Image Generation and Editing System Based on Multi-Turn Dialogue (AAAI 2025 · similarity 0.83)
- Paper2Poster: Towards Multimodal Poster Automation from Scientific Papers (NeurIPS 2025 · similarity 0.83)
- From Words to Structured Visuals: A Benchmark and Framework for Text-to-Diagram Generation and Editing (CVPR 2025 · similarity 0.83)
- Agent-Native Research Artifacts (NeurIPS 2026 · similarity 0.83)
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.
