# Self-Improving CAD Generation Agents with Finite Element Analysis as Feedback

arXiv: https://arxiv.org/abs/2605.17448 · HTML: https://mlsanity.com/paper/2605.17448/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 32 references that are in the map.
- First posted on arXiv 17 May 2026.

## Builds on (most cited first, 22 of 22)

- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [OpenAI GPT-5 System Card](https://mlsanity.com/paper/2601.03267/) (2025)
- [A Point Set Generation Network for 3D Object Reconstruction from a Single Image](https://mlsanity.com/paper/1612.00603/) (2016)
- [DeepCAD: A Deep Generative Network for Computer-Aided Design Models](https://mlsanity.com/paper/2105.09492/) (2021)
- [What Do Single-view 3D Reconstruction Networks Learn?](https://mlsanity.com/paper/1905.03678/) (2019)
- [VideoLLM-online: Online Video Large Language Model for Streaming Video](https://mlsanity.com/paper/2406.11816/) (2024)
- [3D-GPT: Procedural 3D Modeling with Large Language Models](https://mlsanity.com/paper/2310.12945/) (2023)
- [JoinABLe: Learning Bottom-up Assembly of Parametric CAD Joints](https://mlsanity.com/paper/2111.12772/) (2021)
- [CAD-Llama: Leveraging Large Language Models for Computer-Aided Design Parametric 3D Model Generation](https://mlsanity.com/paper/2505.04481/) (2025)
- [CAD-Recode: Reverse Engineering CAD Code from Point Clouds](https://mlsanity.com/paper/2412.14042/) (2024)
- [CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attention](https://mlsanity.com/paper/2402.17678/) (2024)
- [Generating CAD Code with Vision-Language Models for 3D Designs](https://mlsanity.com/paper/2410.05340/) (2024)
- [Text-to-CadQuery: A New Paradigm for CAD Generation with Scalable Large Model Capabilities](https://mlsanity.com/paper/2505.06507/) (2025)
- [CAD-Coder: An Open-Source Vision-Language Model for Computer-Aided Design Code Generation](https://mlsanity.com/paper/2505.14646/) (2025)
- [CAD-Assistant: Tool-Augmented VLLMs as Generic CAD Task Solvers](https://mlsanity.com/paper/2412.13810/) (2024)
- [CADmium: Fine-Tuning Code Language Models for Text-Driven Sequential CAD Design](https://mlsanity.com/paper/2507.09792/) (2025)
- [Seek-CAD: A Self-refined Generative Modeling for 3D Parametric CAD Using Local Inference via DeepSeek](https://mlsanity.com/paper/2505.17702/) (2025)
- [S2O: Static to Openable Enhancement for Articulated 3D Objects](https://mlsanity.com/paper/2409.18896/) (2024)
- [cadrille: Multi-modal CAD Reconstruction with Reinforcement Learning](https://mlsanity.com/paper/2505.22914/) (2025)
- [CADSmith: Multi-Agent CAD Generation with Programmatic Geometric Validation](https://mlsanity.com/paper/2603.26512/) (2026)
- [Drawing2CAD: Sequence-to-Sequence Learning for CAD Generation from Vector Drawings](https://mlsanity.com/paper/2508.18733/) (2025)
- [EvoCAD: Evolutionary CAD Code Generation with Vision Language Models](https://mlsanity.com/paper/2510.11631/) (2025)

## Built on by (most cited first, 2 of 2)

- [CADEngBench: It Looks Like CAD, but Does It Work? Evaluating Parametric Design, Assembly Reasoning, and Physics Simulation](https://mlsanity.com/paper/2608.09296/) (2026)
- [Test-Time Scaling for CAD Generation via Verifier-Free Consensus Selection](https://mlsanity.com/paper/2608.09706/) (2026)

## Abstract (arXiv)

Computer-aided design (CAD) is the backbone of modern industrial design, yet learned CAD generators still fall short of real engineering pipelines: they neither iterate like engineers nor evaluate what engineering requires. Prior work has treated CAD generation as two disjoint steps, part synthesis and assembly, where the former is graded by proximity to a gold reference and the latter, when handled at all, is reduced to a separate constraint solving step. In this work, we introduce a more industry-native task formulation that requires a model to produce a fully assembled multi-part STEP file from a free-form engineering brief, which is then validated via finite element analysis (FEA). FEA validation reveals that Codex (GPT-5.5) and Claude Code (Opus-4.7) agents do not produce a single strict-passing artifact in the main first-attempt sweep, with the best configuration meeting only about 20% of typed requirements on average. Moreover, we introduce two additional supervision signals, a novel text-only blueprint schema and a 21-view image renderer that aids the agent's visual inspection, that better align the generation loop with how engineers iterate in practice. On S2O and Fusion360, the same feedback tools improve geometric reconstruction, with GPT-5.5/xhigh rising from 0.444 to 0.592 Box-IoU on S2O and from 0.397 to 0.505 on Fusion360. Together these signals move CAD programs toward artifacts that are not only visually plausible but also checked against physical and structural requirements.