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

Hugging Face Daily Papers Papers

Summary

This paper introduces a new task formulation for CAD generation that incorporates finite element analysis as feedback, along with improved supervision signals like a text-only blueprint schema and multi-view image renderer, leading to better geometric reconstruction on benchmarks.

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.
Original Article
View Cached Full Text

Cached at: 05/26/26, 02:41 AM

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

Source: https://huggingface.co/papers/2605.17448

Abstract

Learning-based CAD generation systems are enhanced by incorporating engineering validation through finite element analysis and improved supervision signals that better align with actual design processes.

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 treatedCAD generationas 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-partSTEP filefrom a free-form engineering brief, which is then validated viafinite 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 improvegeometric reconstruction, with GPT-5.5/xhigh rising from 0.444 to 0.592Box-IoUon 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.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2605\.17448

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.17448 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.17448 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.17448 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

RA-CAD: Learning Post-Execution Critique for State-Aware Text-to-CAD Generation

arXiv cs.AI

RA-CAD presents a state-aware agent for text-to-CAD generation that uses a Generate–Execute–Critique–Rewrite loop, with feedback-driven agent optimization via Group Relative Policy Optimization. It achieves state-of-the-art execution validity and geometric quality on CADFusion and Text2CAD benchmarks.

Memory-Augmented Reinforcement Learning Agent for CAD Generation

arXiv cs.AI

This paper proposes a memory-augmented reinforcement learning framework for CAD generation agents that integrates geometric kernel toolchains, dual-track memory, and dynamic utility retrieval to handle complex CAD models with long operation sequences and geometric constraints, achieving improved success rate and geometric consistency.