Research map / Vision-language and multimodal / Human and embodied modeling
CADFS: A Big CAD Program Dataset and Framework for Computer-Aided Design with Large Language Models
Vladislav Pyatov, Gleb Bobrovskikh, Saveliy Galochkin, Nikita Boldyrev, Oleg Voynov, Alexander Filippov, Gonzalo Ferrer, Peter Wonka, Evgeny Burnaev
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 new representation-aligned text annotations (ablation) | FeatureScript model, text-to-CAD · DeepCAD test set | CD: 0.40 -> 0.06 | Improved |
| add new representation-aligned text annotations (ablation) | FeatureScript model, text-to-CAD · DeepCAD test set | ECD: 24.4 -> 8.5 | Improved |
| train on all 15 operations instead of sketch and extrude only (ablation) | FeatureScript model, text-to-CAD · Our test set (15 ops) | CD: 14.50 -> 0.07 | Improved |
| train on all 15 operations (ablation) | FeatureScript model, text-to-CAD · DeepCAD test set | CD: 0.06 -> 0.06 | No clear effect |
| train on all 15 operations (ablation) | FeatureScript model, image-to-CAD · Our test set (15 ops) | CD: 8.69 -> 0.35 | Improved |
| train on all 15 operations (ablation) | FeatureScript model, image-to-CAD · Our test set (15 ops) | ECD: 55.6 -> 15.8 | Improved |
| train on all 15 operations (ablation) | FeatureScript model, image-to-CAD · DeepCAD test set | ECD: 17.8 -> 11.5 | Improved |
| train on all 15 operations (ablation) | FeatureScript model, image-to-CAD · Our test set (15 ops) | IR: 22 -> 29 | Hurt |
| replace Python (Cadrille) with FeatureScript on same DeepCAD data and T2C annotations (ablation) | Qwen2-VL-2B, text-to-CAD · DeepCAD test set | ECD: 22.5 -> 24.4 | No clear effect |
| replace Python with FeatureScript on DeepCAD data (ablation) | Qwen2-VL-2B, image-to-CAD · DeepCAD test set | CD / ECD: 0.20 -> 0.14; 14.3 -> 17.8 | Mixed |
| CADFS full framework versus Cadrille | Qwen2-VL-2B, text-to-CAD · DeepCAD test set | CD: 0.10 -> 0.06 | Improved |
| CADFS full framework versus Cadrille | Qwen2-VL-2B, text-to-CAD · DeepCAD test set | IR: 3 -> 9 | Hurt |
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.
- Text2CAD — Text2CAD: Generating Sequential CAD Designs from Beginner-to-Expert Level Text Prompts (NeurIPS 2024) · better on 2 of 2 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
- CAD-Editor: A Locate-then-Infill Framework with Automated Training Data Synthesis for Text-Based CAD Editing (ICML 2025 · similarity 0.90)
- cadrille: Multi-modal CAD Reconstruction with Reinforcement Learning (ICLR 2026 · similarity 0.89)
- From 2D CAD Drawings to 3D Parametric Models: A Vision-Language Approach (AAAI 2025 · similarity 0.89)
- Generating CAD Code with Vision-Language Models for 3D Designs (ICLR 2025 · similarity 0.89)
- CAD-Llama: Leveraging Large Language Models for Computer-Aided Design Parametric 3D Model Generation (CVPR 2025 · similarity 0.89)
- Text2CAD: Generating Sequential CAD Designs from Beginner-to-Expert Level Text Prompts (NeurIPS 2024 · similarity 0.88)
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.
