mlsanityData to 7 Oct 2026

CADReasoner: Iterative Program Editing for CAD Reverse Engineering

arXiv 2603.29847 ↗First posted 18 Feb 2026cs.GR cs.CV cs.HCcited by 4

Not on the map. Why?

Abstract (arXiv)

Computer-Aided Design (CAD) powers modern engineering, yet producing high-quality parts still demands substantial expert effort. Many AI systems tackle CAD reverse engineering, but most are single-pass and miss fine geometric details. In contrast, human engineers compare the input shape with the reconstruction and iteratively modify the design based on remaining discrepancies. Agent-based methods mimic this loop with frozen VLMs, but weak 3D grounding of current foundation models limits reliability and efficiency. We introduce CADReasoner, a model trained to iteratively refine its prediction using geometric discrepancy between the input and the predicted shape. The model outputs a runnable CadQuery Python program whose rendered mesh is fed back at the next step. CADReasoner fuses multi-view renders and point clouds as complementary modalities. To bridge the realism gap, we propose a scan-simulation protocol applied during both training and evaluation. Across DeepCAD, Fusion 360, and MCB benchmarks, CADReasoner attains state-of-the-art results on clean and scan-sim tracks.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Builds on 46 of its 63 references that are in the map. First posted on arXiv 18 Feb 2026.

Builds on

46 of 63 references.

References · 46

  1. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  2. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  3. 2025Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?cited by 948
  4. 2022CodeRL: Mastering Code Generation through Pretrained Models and Deep Reinforcement Learningcited by 342
  5. 2021DeepCAD: A Deep Generative Network for Computer-Aided Design Modelscited by 266
  6. 2025Spurious Rewards: Rethinking Training Signals in RLVRcited by 199
  7. 2018Scan2CAD: Learning CAD Model Alignment in RGB-D Scanscited by 142
  8. 2022SkexGen: Autoregressive Generation of CAD Construction Sequences with Disentangled Codebookscited by 121
  9. 2024BrepGen: A B-rep Generative Diffusion Model with Structured Latent Geometrycited by 101
  10. 2020ParSeNet: A Parametric Surface Fitting Network for 3D Point Cloudscited by 96
  11. 2019Write, Execute, Assess: Program Synthesis with a REPLcited by 92
  12. 2021BRepNet: A topological message passing system for solid modelscited by 92
34 more
  1. 2018Supervised Fitting of Geometric Primitives to 3D Point Cloudscited by 90
  2. 2023Hierarchical Neural Coding for Controllable CAD Model Generationcited by 84
  3. 2022SolidGen: An Autoregressive Model for Direct B-rep Synthesiscited by 78
  4. 2020UCSG-Net -- Unsupervised Discovering of Constructive Solid Geometry Treecited by 77
  5. 2024CAD-MLLM: Unifying Multimodality-Conditioned CAD Generation With MLLMcited by 74
  6. 2019Learning to Infer and Execute 3D Shape Programscited by 68
  7. 2020PIE-NET: Parametric Inference of Point Cloud Edgescited by 67
  8. 2025CAD-Llama: Leveraging Large Language Models for Computer-Aided Design Parametric 3D Model Generationcited by 65
  9. 2024CAD-Recode: Reverse Engineering CAD Code from Point Cloudscited by 63
  10. 2021CAPRI-Net: Learning Compact CAD Shapes with Adaptive Primitive Assemblycited by 62
  11. 2024CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attentioncited by 62
  12. 2024Generating CAD Code with Vision-Language Models for 3D Designscited by 61
  13. 2025CAD-Coder: Text-to-CAD Generation with Chain-of-Thought and Geometric Rewardcited by 53
  14. 2024CAD-GPT: Synthesising CAD Construction Sequence with Spatial Reasoning-Enhanced Multimodal LLMscited by 52
  15. 2021CSG-Stump: A Learning Friendly CSG-Like Representation for Interpretable Shape Parsingcited by 51
  16. 2022Reconstructing editable prismatic CAD from rounded voxel modelscited by 41
  17. 2024CAD-Assistant: Tool-Augmented VLLMs as Generic CAD Task Solverscited by 40
  18. 2023D2CSG: Unsupervised Learning of Compact CSG Trees with Dual Complements and Dropoutscited by 39
  19. 2025CAD-Coder: An Open-Source Vision-Language Model for Computer-Aided Design Code Generationcited by 39
  20. 2022ExtrudeNet: Unsupervised Inverse Sketch-and-Extrude for Shape Parsingcited by 38
  21. 2024Img2CAD: Conditioned 3D CAD Model Generation from Single Image with Structured Visual Geometrycited by 35
  22. 2023Point2CAD: Reverse Engineering CAD Models from 3D Point Cloudscited by 33
  23. 2024FlexCAD: Unified and Versatile Controllable CAD Generation with Fine-tuned Large Language Modelscited by 33
  24. 2024Query2CAD: Generating CAD models using natural language queriescited by 28
  25. 2025CADDreamer: CAD Object Generation from Single-view Imagescited by 22
  26. 2025Seek-CAD: A Self-refined Generative Modeling for 3D Parametric CAD Using Local Inference via DeepSeekcited by 22
  27. 2021PvDeConv: Point-Voxel Deconvolution for Autoencoding CAD Construction in 3Dcited by 20
  28. 2023SHARP Challenge 2023: Solving CAD History and pArameters Recovery from Point clouds and 3D scans. Overview, Datasets, Metrics, and Baselinescited by 19
  29. 2025CAD-Editor: A Locate-then-Infill Framework with Automated Training Data Synthesis for Text-Based CAD Editingcited by 19
  30. 2022Neural Face Identification in a 2D Wireframe Projection of a Manifold Objectcited by 18
  31. 2024Split-and-Fit: Learning B-Reps via Structure-Aware Voronoi Partitioningcited by 17
  32. 2025CAD-Coder:Text-Guided CAD Files Code Generationcited by 8
  33. 2024Learning to Edit Visual Programs with Self-Supervisioncited by 7
  34. 2025GACO-CAD: Geometry-Augmented and Conciseness-Optimized CAD Model Generation from Single Imagecited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026IterCAD: An Iterative Multimodal Agent for Visually-Grounded CAD Generation and Editingcited by 9
  2. 2026CADWorld: Computer-Use Benchmark for Long-Horizon Computer-Aided Designcited by 1
  3. 2026CADENA: Stepwise CAD Reverse Engineeringcited by 0
  4. 2026FACT: Fidelity-Aware Construction of Articulated Twinscited by 0