Research map / Graph learning / Graph self-supervised learning
Learning Hierarchical and Geometry-Aware Graph Representations for Text-to-CAD
Shengjie Gong, Wenjie Peng, Hongyuan Chen, Gangyu Zhang, Yunqing Hu, Huiyuan Zhang, Shuangping Huang, Tianshui Chen
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 graph decomposition stage (full three-stage vs w/o Graph Decom.) (ablation) | Graph-CAD (SFT) · CADBench-Sim | GCS: 0.7268 -> 0.7830 | Improved |
| add action planning stage (vs w/o Action Planning) (ablation) | Graph-CAD (SFT) · CADBench-Wild | syntax error rate: 11.0% -> 4.5% | Improved |
| add action planning stage (vs w/o Action Planning) (ablation) | Graph-CAD (SFT) · CADBench-Sim | Inst.: 0.4006 -> 0.4733 | Improved |
| add both graph and action stages (vs end-to-end) (ablation) | Graph-CAD (SFT) · CADBench-Sim | Avg.: 0.5573 -> 0.6431 | Improved |
| add both stages (vs end-to-end) (ablation) | Graph-CAD (SFT) · inference time | seconds per sample: 64.861 -> 104.755 | Hurt |
| three-stage graph inference with two-shot prompts | GPT-5 · CADBench-Sim | GCS: 0.3846 -> 0.6603 | Improved |
| three-stage graph inference with two-shot prompts | GPT-5 · CADBench-Wild | GCS: 0.4017 -> 0.5849 | Improved |
| three-stage graph inference with two-shot prompts | Claude-opus-4-1 · CADBench-Sim | GCS: 0.4932 -> 0.5705 | Improved |
| three-stage graph inference with two-shot prompts | Claude-opus-4-1 · CADBench-Sim | Inst.: 0.5403 -> 0.5025 | Hurt |
| three-stage graph inference with two-shot prompts | Claude-opus-4-1 · CADBench-Sim | Avg.: 0.6662 -> 0.6664 | No clear effect |
| add SAPCL curriculum learning on top of SFT | Graph-CAD · CADBench-Sim | GCS: 0.7830 -> 0.9018 | Improved |
| add SAPCL curriculum learning on top of SFT | Graph-CAD · CADBench-Wild | Avg.: 0.6692 -> 0.7114 | Improved |
Closest papers by meaning
- CAD-Coder: Text-to-CAD Generation with Chain-of-Thought and Geometric Reward (NeurIPS 2025 · similarity 0.89)
- StepCAD: Mesh-to-CAD Code Generation via LLM Policy and Geometry-Guided Search (NeurIPS 2026 · similarity 0.87)
- CAD-Editor: A Locate-then-Infill Framework with Automated Training Data Synthesis for Text-Based CAD Editing (ICML 2025 · similarity 0.87)
- Clarify Before You Draw: Proactive Agents for Robust Text-to-CAD Generation (ICML 2026 · similarity 0.87)
- FlexCAD: Unified and Versatile Controllable CAD Generation with Fine-tuned Large Language Models (ICLR 2025 · similarity 0.86)
- Text2CAD: Generating Sequential CAD Designs from Beginner-to-Expert Level Text Prompts (NeurIPS 2024 · similarity 0.86)
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.
