pith. machine review for the scientific record. sign in

arxiv: 2604.15184 · v2 · submitted 2026-04-16 · 💻 cs.AI

Recognition: unknown

Agent-Aided Design for Dynamic CAD Models

Mitch Adler , Matthew Russo , Michael Cafarella
Authors on Pith no claims yet

Pith reviewed 2026-05-10 11:09 UTC · model grok-4.3

classification 💻 cs.AI
keywords agent-aided designdynamic CAD3D assembliesmovable partsconstraint solversLLM agentsFreeCADdegrees of freedom
0
0 comments X

The pith

An agentic system can generate 3D CAD assemblies with moving parts by using external solvers and visual feedback.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Existing agent-aided design systems create static CAD objects but cannot produce assemblies that move, such as a piston, pendulum, or scissors. This paper presents AADvark, a prototype that lets an agent build models while directly capturing interactions with one or more degrees of freedom. The system modifies the agent's CAD tool and assembly solver to supply a reliable verification signal inside an iterative loop. A reader would care because functional moving parts are required for most real manufacturing applications.

Core claim

AADvark captures dynamic part interactions with one or more degrees-of-freedom, allowing the agent to reason directly about assemblies with moving parts. External constraint solver tools combined with specialized visual feedback create a strong verification signal that enables the system to produce valid dynamic 3D assemblies despite imperfect LLM spatial reasoning.

What carries the argument

The iterative agent loop in AADvark that modifies FreeCAD and the assembly solver to generate and check dynamic CAD models with degrees of freedom.

Load-bearing premise

Large language models stay imperfect at spatial reasoning but can still reach correct dynamic assemblies when supplied with external constraint solvers and visual feedback.

What would settle it

A simple test case such as a pair of scissors where the system produces no valid moving assembly after repeated iterations with the modified solver and visual checks.

Figures

Figures reproduced from arXiv: 2604.15184 by Matthew Russo, Michael Cafarella, Mitch Adler.

Figure 1
Figure 1. Figure 1: An illustration of AADvark generating a pair of scissors. AADvark accepts one or more images and/or textual description as input. AADvark creates JSON definitions for the parts and joints in the 3D assembly and then compiles them using a 3D assembly constraint solver. Compilation errors and intermediate renderings (generated in FreeCAD) are fed back into the agent. We modify FreeCAD and the constraint solv… view at source ↗
Figure 2
Figure 2. Figure 2: Snapshots from our demonstration of AADvark creating a dynamic 3D assembly for a pair of scissors. Each column corresponds to a single iteration of the object as it was generated. Each row shows the object at a different angle of the revolute joint (0, 20, 40, and 60 degrees). AADvark starts out with a base rectangle and by the end of our demonstration it has created a functional pair of scissors. joint to… view at source ↗
read the original abstract

In the past year, researchers have created agentic systems that can design real-world CAD-style objects in a training-free setting, a new variety of system that we call Agent-Aided Design. These systems place an agent in a feedback loop in which it generates an assembly of CAD model(s), visualizes the assembly, and then iteratively refines its assembly based on visual and other feedback. Despite rapid progress, a key problem remains: none of these systems can build complex 3D assemblies with moving parts. For example, no existing system can build a piston, a pendulum, or even a pair of scissors. In order for Agent-Aided Design to make a real impact in industrial manufacturing, we need a system that is capable of generating such 3D assemblies. In this paper we present a prototype of AADvark, an agentic system designed for this task. Unlike previous state-of-the-art systems, AADvark captures the dynamic part interactions with one or more degrees-of-freedom. This design decision allows AADvark to reason directly about assemblies with moving parts and can thereby achieve cross-cutting goals, including but not limited to mechanical movements. Unfortunately, current LLMs are imperfect spatial reasoners, a problem that AADvark addresses by incorporating external constraint solver tools with a specialized visual feedback mechanism. We demonstrate that, by modifying the agent's tools (FreeCAD and the assembly solver), we are able to create a strong verification signal which enables our system to build 3D assemblies with movable parts.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper introduces AADvark, a prototype agentic system for Agent-Aided Design that generates 3D CAD assemblies with movable parts (e.g., pistons, pendulums, scissors) by capturing dynamic interactions with one or more degrees of freedom. It places an LLM agent in an iterative loop that generates assemblies, visualizes them, and refines based on feedback, addressing LLM spatial reasoning limitations via modified FreeCAD and assembly solver tools that purportedly produce a strong verification signal.

Significance. If the core demonstration holds, the work would extend agentic CAD systems beyond static objects to dynamic mechanisms, with potential industrial relevance. The use of external constraint solvers and specialized visual feedback to compensate for imperfect LLM reasoning is a constructive direction, but the complete absence of quantitative results, metrics, or baselines prevents any assessment of whether the claimed verification signal is actually strong or effective.

major comments (1)
  1. [Abstract] Abstract: the central claim that tool modifications to FreeCAD and the assembly solver 'create a strong verification signal which enables our system to build 3D assemblies with movable parts' is unsupported. The manuscript supplies no definition of signal strength, no success rates, no iteration statistics, no failure-mode analysis, and no unmodified-baseline comparison, rendering it impossible to evaluate whether the iterative loop produces valid dynamic constraints or simply masks spatial errors.
minor comments (1)
  1. The terms 'strong verification signal' and 'dynamic part interactions' are introduced without operational definitions or examples, which hinders readability even if the empirical gap is addressed.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback. We agree that the central claim in the abstract requires quantitative support and will revise the manuscript accordingly to include metrics, statistics, and baseline comparisons.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that tool modifications to FreeCAD and the assembly solver 'create a strong verification signal which enables our system to build 3D assemblies with movable parts' is unsupported. The manuscript supplies no definition of signal strength, no success rates, no iteration statistics, no failure-mode analysis, and no unmodified-baseline comparison, rendering it impossible to evaluate whether the iterative loop produces valid dynamic constraints or simply masks spatial errors.

    Authors: We acknowledge this limitation in the current manuscript, which presents the system as a prototype through qualitative demonstrations of assemblies such as pistons, pendulums, and scissors. We will add a dedicated quantitative evaluation section in the revision. This section will define verification signal strength as the fraction of trials in which the agent produces a fully constraint-satisfying assembly within a maximum of 10 iterations. We will report success rates, mean and variance of iteration counts, and a categorized failure-mode analysis (e.g., constraint solver rejections versus visualization-detected spatial errors) across 30 independent runs per mechanism. We will also include a baseline comparison using unmodified FreeCAD and solver interfaces without the specialized visual feedback loop, quantifying the improvement attributable to our tool modifications. These additions will directly address the lack of metrics and enable evaluation of the claimed verification signal. revision: yes

Circularity Check

0 steps flagged

No circularity: descriptive system paper with external tools and no self-referential derivations

full rationale

The paper presents a prototype agentic system (AADvark) that integrates LLMs with modified external tools (FreeCAD and assembly solver) to generate dynamic 3D assemblies. No equations, fitted parameters, predictions, or first-principles derivations appear in the provided text. The central claim—that tool modifications create a 'strong verification signal'—is a descriptive assertion about system behavior rather than a reduction of outputs to inputs by construction. No self-citations, ansatzes, or uniqueness theorems are invoked as load-bearing elements. The methodology relies on independent external components and iterative feedback loops, making the approach self-contained without circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the system description centers on tool integration rather than new theoretical constructs.

pith-pipeline@v0.9.0 · 5566 in / 1023 out tokens · 47410 ms · 2026-05-10T11:09:41.411536+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 23 canonical work pages · 5 internal anchors

  1. [1]

    GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

    Lakshya A Agrawal, Shangyin Tan, Dilara Soylu, Noah Ziems, Rishi Khare, Krista Opsahl-Ong, Arnav Singhvi, Herumb Shandilya, Michael J Ryan, Meng Jiang, Christopher Potts, Koushik Sen, Alexandros G. Dimakis, Ion Stoica, Dan Klein, Matei Zaharia, and Omar Khattab. 2025. GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning. arXiv:2507.1945...

    work page internal anchor Pith review arXiv 2025

  2. [2]

    Kamel Alrashedy, Pradyumna Tambwekar, Zulfiqar Haider Zaidi, Megan Lang- wasser, Wei Xu, and Matthew Gombolay. 2025. Generating CAD Code with Vision- Language Models for 3D Designs. InThe Thirteenth International Conference on Learning Representations. https://openreview.net/forum?id=BLWaTeucYX

    2025

  3. [3]

    Cheng Chen, Jiacheng Wei, Tianrun Chen, Chi Zhang, Xiaofeng Yang, Shangzhan Zhang, Bingchen Yang, Chuan-Sheng Foo, Guosheng Lin, Qixing Huang, and Fayao Liu. 2025. CADCrafter: Generating Computer-Aided Design Models from Unconstrained Images. In2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 11073–11082. doi:10.1109/CVPR52734.2...

    work page doi:10.1109/cvpr52734.2025.01034 2025

  4. [4]

    Audrey Cheng, Shu Liu, Melissa Pan, Zhifei Li, Shubham Agarwal, Mert Cemri, Bowen Wang, Alexander Krentsel, Tian Xia, Jongseok Park, Shuo Yang, Jeff Chen, Lakshya Agrawal, Ashwin Naren, Shulu Li, Ruiying Ma, Aditya Desai, Jiarong Xing, Koushik Sen, Matei Zaharia, and Ion Stoica. 2025. Let the Barbarians In: How AI Can Accelerate Systems Performance Resear...

    work page arXiv 2025

  5. [5]

    Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. Training Verifiers to Solve Math Word Problems. arXiv:2110.14168 [cs.LG] https://arxiv.org/abs/2110.14168

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  6. [6]

    2026.FreeCAD

    FreeCAD contributors. 2026.FreeCAD. https://github.com/FreeCAD/FreeCAD

    2026

  7. [7]

    2026.OndselSolver

    OndselSolver contributors and Mitch Adler. 2026.OndselSolver. https://github. com/FreeCAD/OndselSolver

    2026

  8. [8]

    Yuanzhe Deng, James Chen, and Alison Olechowski. 2023. What Sets Profi- cient and Expert Users Apart? Results of a Computer-Aided Design Experiment. Journal of Mechanical Design146 (09 2023), 1–39. doi:10.1115/1.4063360

    work page doi:10.1115/1.4063360 2023

  9. [9]

    2024.IKEA’s Journey Through 3D Visualization and Spatial Computing

    Martin Enthed and AWE XR. 2024.IKEA’s Journey Through 3D Visualization and Spatial Computing. Retrieved January 29, 2026 from https://www.youtube.com/ watch?v=Kup0d4Te3n0

    2024

  10. [10]

    Fengxiao Fan, Jingzhe Ni, Xiaolong Yin, Sirui Wang, Xingyu Lu, Qiang Zou, Ruofeng Tong, Min Tang, and Peng Du. 2025. CADDesigner: Conceptual Design of CAD Models Based on General-Purpose Agent. arXiv:2508.01031 [cs.AI] https://arxiv.org/abs/2508.01031

    work page internal anchor Pith review arXiv 2025

  11. [11]

    Pouya Hamadanian, Pantea Karimi, Arash Nasr-Esfahany, Kimia Noorbakhsh, Joseph Chandler, Ali ParandehGheibi, Mohammad Alizadeh, and Hari Balakr- ishnan. 2025. Glia: A Human-Inspired AI for Automated Systems Design and Optimization. arXiv:2510.27176 [cs.AI] https://arxiv.org/abs/2510.27176

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  12. [12]

    Carlos E Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik R Narasimhan. 2024. SWE-bench: Can Language Models Resolve Real-world Github Issues?. InThe Twelfth International Conference on Learning Representations. https://openreview.net/forum?id=VTF8yNQM66

    2024

  13. [13]

    Mohammad Sadil Khan, Elona Dupont, Sk Aziz Ali, Kseniya Cherenkova, Anis Kacem, and Djamila Aouada. 2024. CAD-SIGNet: CAD Language Inference from Point Clouds using Layer-wise Sketch Instance Guided Attention. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 4713–4722

    2024

  14. [14]

    Mohammad Sadil Khan, Sankalp Sinha, Sheikh Talha Uddin, Didier Stricker, Sk Aziz Ali, and Muhammad Zeshan Afzal. 2024. Text2CAD: Generating Se- quential CAD Designs from Beginner-to-Expert Level Text Prompts. InAd- vances in Neural Information Processing Systems, Vol. 37. Curran Associates, Inc., 7552–7579. https://proceedings.neurips.cc/paper_files/paper...

    2024

  15. [15]

    Joshi, Hanna Moazam, Heather Miller, Matei Zaharia, and Christopher Potts

    Omar Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav Santhanam, Sri Vardhamanan, Saiful Haq, Ashutosh Sharma, Thomas T. Joshi, Hanna Moazam, Heather Miller, Matei Zaharia, and Christopher Potts. 2024. DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines. The Twelfth International Conference on Learning Representations

    2024

  16. [16]

    2021 , url =

    Joseph G. Lambourne, Karl D. D. Willis, Pradeep Kumar Jayaraman, Aditya Sanghi, Peter Meltzer, and Hooman Shayani. 2021. BRepNet: A topological message passing system for solid models. In2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 12768–12777. doi:10.1109/CVPR46437.2021.01258

    work page doi:10.1109/cvpr46437.2021.01258 2021

  17. [17]

    Robert Tjarko Lange, Yuki Imajuku, and Edoardo Cetin. 2026. ShinkaEvolve: Towards Open-Ended and Sample-Efficient Program Evolution. InThe Fourteenth International Conference on Learning Representations. https://openreview.net/ forum?id=lKEdGCoDNC

    2026

  18. [18]

    Jiahao Li, Weijian Ma, Xueyang Li, Yunzhong Lou, Guichun Zhou, and Xiangdong Zhou. 2025. CAD-Llama: Leveraging Large Language Models for Computer- Aided Design Parametric 3D Model Generation. In2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 18563–18573. doi:10.1109/ CVPR52734.2025.01730

    work page arXiv 2025

  19. [19]

    Xueyang Li, Jiahao Li, Yu Song, Yunzhong Lou, and Xiangdong Zhou. 2026. Seek- CAD: A Self-refined Generative Modeling for 3D Parametric CAD Using Local Inference via DeepSeek. InThe Fourteenth International Conference on Learning Representations. https://openreview.net/forum?id=PzIc2TxhwN

    2026

  20. [20]

    Paweł Liskowski, Benjamin Han, Paritosh Aggarwal, Bowei Chen, Boxin Jiang, Nitish Jindal, Zihan Li, Aaron Lin, Kyle Schmaus, Jay Tayade, Weicheng Zhao, Anupam Datta, Nathan Wiegand, and Dimitris Tsirogiannis. 2025. Cortex AISQL: A Production SQL Engine for Unstructured Data. arXiv:2511.07663 [cs.DB] https://arxiv.org/abs/2511.07663

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  21. [21]

    Chunwei Liu, Matthew Russo, Michael Cafarella, Lei Cao, Peter Baile Chen, Zui Chen, Michael Franklin, Tim Kraska, Samuel Madden, Rana Shahout, et al. 2025. Palimpzest: Optimizing AI-Powered Analytics with Declarative Query Processing. CIDR

    2025

  22. [22]

    Dimitrios Mallis, Ahmet Serdar Karadeniz, Sebastian Cavada, Danila Rukhovich, Niki Maria Foteinopoulou, Kseniya Cherenkova, Anis Kacem, and Djamila Aouada. 2025. CAD-Assistant: Tool-Augmented VLLMs as Generic CAD Task Solvers.International Conference on Computer Vision (ICCV)(2025)

    2025

  23. [23]

    2026.OnShape

    OnShape. 2026.OnShape. https://www.onshape.com/en/

    2026

  24. [24]

    Liana Patel, Siddharth Jha, Melissa Pan, Harshit Gupta, Parth Asawa, Carlos Guestrin, and Matei Zaharia. 2025. Semantic Operators and Their Optimization: Enabling LLM-Based Data Processing with Accuracy Guarantees in LOTUS.Proc. VLDB Endow.18, 11 (July 2025), 4171–4184. doi:10.14778/3749646.3749685

    work page doi:10.14778/3749646.3749685 2025

  25. [25]

    Robertson and T.J

    D. Robertson and T.J. Allen. 1993. CAD system use and engineering performance. IEEE Transactions on Engineering Management40, 3 (1993), 274–282. doi:10.1109/ 17.233189

    1993

  26. [26]

    Danila Rukhovich, Elona Dupont, Dimitrios Mallis, Kseniya Cherenkova, Anis Kacem, and Djamila Aouada. 2025. CAD-Recode: Reverse Engineering CAD Code from Point Clouds. arXiv:2412.14042 [cs.CV] https://arxiv.org/abs/2412.14042

    work page arXiv 2025

  27. [27]

    Matthew Russo and Tim Kraska. 2025. Deep Research is the New An- alytics System: Towards Building the Runtime for AI-Driven Analytics. arXiv:2509.02751 [cs.AI] https://arxiv.org/abs/2509.02751

    work page arXiv 2025

  28. [28]

    Matthew Russo, Sivaprasad Sudhir, Gerardo Vitagliano, Chunwei Liu, Tim Kraska, Samuel Madden, and Michael Cafarella. 2025. Abacus: A Cost-Based Optimizer for Semantic Operator Systems. arXiv:2505.14661 [cs.DB] https://arxiv.org/abs/ 2505.14661

    work page arXiv 2025

  29. [29]

    Shankar et al

    Shreya Shankar, Tristan Chambers, Tarak Shah, Aditya G. Parameswaran, and Eugene Wu. 2025. DocETL: Agentic Query Rewriting and Evaluation for Complex Document Processing.Proc. VLDB Endow.18, 9 (May 2025), 3035–3048. doi:10. 14778/3746405.3746426

    work page arXiv 2025

  30. [30]

    2026.SolidWorks Design

    SolidWorks. 2026.SolidWorks Design. https://www.solidworks.com/

    2026

  31. [31]

    2023.Designing a winning Formula 1 car with NX CAD | Oracle Red Bull Racing

    Alex Stoermann. 2023.Designing a winning Formula 1 car with NX CAD | Oracle Red Bull Racing. Retrieved January 29, 2026 from https://www.swooshtech.com/ 2023/03/24/designing-a-winning-formula-1-car-with-nx-cad

    2023

  32. [32]

    Tsaftaris

    Ilias Stogiannidis, Steven McDonagh, and Sotirios A. Tsaftaris. 2025. Mind the Gap: Benchmarking Spatial Reasoning in Vision-Language Models. arXiv:2503.19707 [cs.CV] https://arxiv.org/abs/2503.19707

    work page arXiv 2025

  33. [33]

    Siyu Wang, Cailian Chen, Xinyi Le, Qimin Xu, Lei Xu, Yanzhou Zhang, and Jie Yang. 2025. CAD-GPT: synthesising CAD construction sequence with spatial reasoning-enhanced multimodal LLMs. InProceedings of the Thirty-Ninth AAAI Conference on Artificial Intelligence and Thirty-Seventh Conference on Innovative Applications of Artificial Intelligence and Fifteen...

    work page doi:10.1609/aaai.v39i8.32849 2025

  34. [34]

    Parameswaran

    Lindsey Linxi Wei, Shreya Shankar, Sepanta Zeighami, Yeounoh Chung, Fatma Ozcan, and Aditya G. Parameswaran. 2026. Multi-Objective Agentic Rewrites for Unstructured Data Processing. arXiv:2512.02289 [cs.DB] https://arxiv.org/abs/ 2512.02289

    work page arXiv 2026

  35. [35]

    Karl D. D. Willis, Yewen Pu, Jieliang Luo, Hang Chu, Tao Du, Joseph G. Lambourne, Armando Solar-Lezama, and Wojciech Matusik. 2021. Fusion 360 gallery: a dataset and environment for programmatic CAD construction from human design sequences.ACM Trans. Graph.40, 4, Article 54 (July 2021), 24 pages. doi:10.1145/ 3450626.3459818

    work page arXiv 2021

  36. [36]

    Rundi Wu, Chang Xiao, and Changxi Zheng. 2021. DeepCAD: A Deep Generative Network for Computer-Aided Design Models. In2021 IEEE/CVF International Conference on Computer Vision (ICCV). 6752–6762. doi:10.1109/ICCV48922.2021. 00670

    work page doi:10.1109/iccv48922.2021 2021

  37. [37]

    Haoyang Xie and Feng Ju. 2025. Text-to-CadQuery: A New Paradigm for CAD Generation with Scalable Large Model Capabilities. arXiv:2505.06507 [cs.AI] https://arxiv.org/abs/2505.06507

    work page arXiv 2025

  38. [38]

    Jingwei Xu, Chenyu Wang, Zibo Zhao, Wen Liu, Yi Ma, and Shenghua Gao. 2025. CAD-MLLM: Unifying Multimodality-Conditioned CAD Generation With MLLM. arXiv:2411.04954 [cs.CV] https://arxiv.org/abs/2411.04954

    work page arXiv 2025

  39. [39]

    Yang You, Mikaela Angelina Uy, Jiaqi Han, Rahul Thomas, Haotong Zhang, Yi Du, Hansheng Chen, Francis Engelmann, Suya You, and Leonidas Guibas

  40. [40]

    InProceedings of the SIGGRAPH Asia 2025 Conference Papers (SA Conference Papers ’25)

    Img2CAD: Reverse Engineering 3D CAD Models from Images through VLM-Assisted Conditional Factorization. InProceedings of the SIGGRAPH Asia 2025 Conference Papers (SA Conference Papers ’25). Association for Computing Machinery, New York, NY, USA, Article 95, 12 pages. doi:10.1145/3757377.3763891

    work page doi:10.1145/3757377.3763891 2025