pith. sign in

arxiv: 2605.19717 · v1 · pith:CTQLBZU4new · submitted 2026-05-19 · 💻 cs.CV

Physics-in-the-Loop: A Hybrid Agentic Architecture for Validated CAD Engineering Design

Elias Berger , Muhammad Usama , Jan Mehlst\"aubl , Bernhard Saske , Kristin Paetzold-Byhain This is my paper

Pith reviewed 2026-05-20 05:52 UTC · model grok-4.3

classification 💻 cs.CV
keywords CAD generationagentic systemsphysical verificationhybrid architecturegenerative designknowledge-based engineeringfunctional validity
0
0 comments X

The pith

Embedding physical verification tools into AI agent loops enables more complex and physically valid CAD designs.

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

The paper argues that large language models alone cannot generate reliable engineering CAD because they lack built-in physical comprehension. To fix this, it places knowledge-based engineering tools inside the agents' decision process so that designs are planned, created, checked for physical validity, and revised in a repeating loop. The authors test this on a new benchmark and report designs that are both more structurally complex and more likely to compile correctly. If the approach holds, automated design systems could move from producing plausible-looking shapes to producing shapes that satisfy real mechanical constraints with less human intervention.

Core claim

Engineering design is cast as a closed-loop sequential decision process in which dedicated agents iteratively plan, generate, evaluate, and revise CAD models, using explicit physical verification signals from knowledge-based engineering tools as the guiding feedback; this hybrid setup yields designs with greater structural complexity and higher functional validity than pure agentic baselines.

What carries the argument

The Hybrid Agentic-Physical Architecture, which inserts validated knowledge-based engineering tools directly into the agents' sequential decision loop to supply physical verification as an explicit feedback signal.

Load-bearing premise

The selected knowledge-based engineering tools supply sufficiently complete and accurate physical verification signals across the full range of designs the agents produce.

What would settle it

An experiment that removes the physical verification loop while keeping all other agent components fixed and measures no difference in structural complexity or compile rate would falsify the claimed benefit of the hybrid architecture.

Figures

Figures reproduced from arXiv: 2605.19717 by Bernhard Saske, Elias Berger, Jan Mehlst\"aubl, Kristin Paetzold-Byhain, Muhammad Usama.

Figure 1
Figure 1. Figure 1: Generated CAD design examples. Top row: Input load [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Condensed excerpt of a load case JSON def [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hybrid Agentic-Physical Architecture. The system processes structured load cases (left) through a multi-agent generation loop. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Generated CAD objects of the Hybrid Agentic Architecture. The top row shows visualizations of the input load case (red: forces, [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Each row illustrates the iterative refinement of CAD mod [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative comparison showing the fidelity of (a) Cad [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
read the original abstract

Large Language Models (LLMs) can generate Computer-Aided Design (CAD), yet lack physical comprehension required for reliable engineering design. Instead of attempting to implicitly learn physical laws from data, we propose a Hybrid Agentic-Physical Architecture that embeds validated knowledge-based engineering tools directly into the decision making loop of autonomous AI agents. In this framework, engineering design is formulated as a closed-loop, sequential decision making process guided by explicit physical verification. Based on a load case, dedicated agents iteratively plan, generate, evaluate, and revise engineering designs using knowledge-based tools as a feedback signal. We introduce a benchmark dataset and metrics for assessing functional validity in generative CAD. Our system generates more complex and physically verified designs, with a 4.2 increase in structural complexity and improving compile rate by 3.5% compared to similar agentic methods. The codebase, prompts and dataset will be made publicly available to support reproducibility and future research.

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

2 major / 2 minor

Summary. The paper proposes a Hybrid Agentic-Physical Architecture that embeds validated knowledge-based engineering (KBE) tools directly into the decision-making loop of autonomous LLM agents for CAD design. Engineering design is cast as a closed-loop sequential process in which dedicated agents iteratively plan, generate, evaluate, and revise designs using explicit physical verification signals from KBE tools. The authors introduce a benchmark dataset and associated metrics for functional validity in generative CAD, and report that the system produces designs with a 4.2-fold increase in structural complexity and a 3.5% higher compile rate relative to comparable agentic baselines.

Significance. If the empirical comparisons and physical-verification claims hold under detailed scrutiny, the work would offer a concrete route to more reliable generative engineering design by replacing implicit physics learning with explicit, tool-based feedback. The planned public release of code, prompts, and dataset would further strengthen reproducibility and enable follow-on research in agentic systems for CAD.

major comments (2)
  1. [Abstract] Abstract: the quantitative claims of a 4.2 increase in structural complexity and 3.5% compile-rate improvement are presented without any accompanying methodology details, error bars, dataset statistics, ablation results, or baseline descriptions, rendering the central empirical result unevaluable from the supplied text.
  2. [Framework paragraph] Framework paragraph: the assertion that designs are 'physically verified' rests on the assumption that the embedded KBE tools supply sufficiently complete signals across the full range of generated designs; the manuscript does not specify the scope of modeled phenomena (e.g., whether thermal effects, fatigue, or contact dynamics are included beyond the benchmark static load cases), which directly affects the interpretation of both the complexity gain and the compile-rate improvement.
minor comments (2)
  1. Clarify the exact definition of 'structural complexity' metric and how it is computed from the CAD outputs.
  2. Add a table or section summarizing the benchmark dataset (number of designs, load-case distribution, and split statistics).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. We address each major comment below and describe the revisions we will implement to improve clarity and rigor.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the quantitative claims of a 4.2 increase in structural complexity and 3.5% compile-rate improvement are presented without any accompanying methodology details, error bars, dataset statistics, ablation results, or baseline descriptions, rendering the central empirical result unevaluable from the supplied text.

    Authors: We agree that the abstract should be more self-contained to allow evaluation of the central claims. The full manuscript provides the requested details, including the benchmark dataset construction, baseline agentic architectures, evaluation metrics for functional validity, ablation studies, and statistical reporting with error bars, in Sections 4 and 5. To directly address the concern, we will revise the abstract to incorporate a concise description of the experimental protocol, the specific baselines used, and the evaluation methodology. This change will make the quantitative results more readily evaluable from the abstract while preserving its brevity. revision: yes

  2. Referee: [Framework paragraph] Framework paragraph: the assertion that designs are 'physically verified' rests on the assumption that the embedded KBE tools supply sufficiently complete signals across the full range of generated designs; the manuscript does not specify the scope of modeled phenomena (e.g., whether thermal effects, fatigue, or contact dynamics are included beyond the benchmark static load cases), which directly affects the interpretation of both the complexity gain and the compile-rate improvement.

    Authors: We acknowledge that explicitly delineating the scope of the KBE verification is necessary for accurate interpretation of the results. The current tools implement static structural analysis under the benchmark load cases, covering stress distribution, deformation limits, and basic geometric compliance. Thermal effects, fatigue, and contact dynamics are outside the modeled scope for this benchmark, which targets functional validity for static load-bearing structures. We will add a new subsection to the Framework section that precisely describes the modeled physical phenomena, lists the verification signals provided to the agents, and discusses the resulting limitations on the generality of the 'physically verified' claim. This addition will clarify how the reported gains in complexity and compile rate should be understood. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical claims rest on independent benchmark comparisons

full rationale

The paper proposes a hybrid agentic architecture that embeds KBE tools into LLM agents for iterative CAD design with explicit physical verification feedback. The central claims are empirical results on a new benchmark dataset: a 4.2 increase in structural complexity and 3.5% compile-rate improvement versus other agentic methods. No equations, first-principles derivations, fitted parameters, or predictions appear in the abstract or described framework. The closed-loop process is a methodological description, not a self-referential mathematical reduction. No self-citations, uniqueness theorems, or ansatzes are invoked to justify core results. The evaluation metrics and dataset are presented as external to the method, making the derivation chain self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are stated. The architecture implicitly assumes that existing knowledge-based engineering tools can serve as reliable, general-purpose physical oracles.

axioms (1)
  • domain assumption Knowledge-based engineering tools provide accurate and sufficient physical verification for generated CAD designs
    The closed-loop feedback mechanism depends on these tools being trustworthy oracles.

pith-pipeline@v0.9.0 · 5707 in / 1131 out tokens · 32459 ms · 2026-05-20T05:52:56.175463+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

90 extracted references · 90 canonical work pages · 12 internal anchors

  1. [1]

    Engineering Design: A Systematic Approach , author =

    work page

  2. [2]

    A Survey on Generative

    Zhang, Wei and Chen, Wei and Zhao, Liang , date =. A Survey on Generative. 2025 , journal =

    work page 2025

  3. [3]

    Challenges and Opportunities in the Integration of Generative

    Berger, Elias and Dammann, Maximilian Peter and Mehlst. Challenges and Opportunities in the Integration of Generative. Proceedings of the Design Society , volume =

    work page

  4. [4]

    Enhancing Computer-Aided Design with Deep Learning Frameworks: A Literature Review , shorttitle =

    Steininger, Sarah and Zhao, Jasmin and Fottner, Johannes , year = 2025, month = aug, journal =. Enhancing Computer-Aided Design with Deep Learning Frameworks: A Literature Review , shorttitle =. doi:10.1017/pds.2025.10165 , urldate =

    work page doi:10.1017/pds.2025.10165 2025

  5. [5]

    2026 , note =

    From Geometry to Function: Towards Context-Aware Generative AI for Engineering Design , author =. 2026 , note =

    work page 2026

  6. [6]

    Cad-mllm: Unifying multimodality-conditioned cad generation with mllm

    Xu, Jingwei and Zhao, Zibo and Wang, Chenyu and Liu, Wen and Ma, Yi and Gao, Shenghua , year = 2025, month = mar, number =. doi:10.48550/arXiv.2411.04954 , urldate =. arXiv , langid =:2411.04954 , primaryclass =

    work page doi:10.48550/arxiv.2411.04954 2025

  7. [7]

    arXiv , howpublished =:2510.11631 , primaryclass =

    Preintner, Tobias and Yuan, Weixuan and K. arXiv , howpublished =:2510.11631 , primaryclass =

    work page arXiv

  8. [8]

    Generating CAD code with vision-language models for 3d designs.arXiv preprint arXiv:2410.05340, 2024

    Alrashedy, Kamel and Tambwekar, Pradyumna and Zaidi, Zulfiqar and Langwasser, Megan and Xu, Wei and Gombolay, Matthew , year = 2025, month = feb, number =. Generating. doi:10.48550/arXiv.2410.05340 , urldate =. arXiv , langid =:2410.05340 , primaryclass =

    work page doi:10.48550/arxiv.2410.05340 2025

  9. [9]

    doi:10.48550/arXiv.2505.08686 , urldate =

    He, Changqi and Zhang, Shuhan and Zhang, Liguo and Miao, Jiajun , year = 2025, month = may, number =. doi:10.48550/arXiv.2505.08686 , urldate =. arXiv , langid =:2505.08686 , primaryclass =

    work page doi:10.48550/arxiv.2505.08686 2025

  10. [10]

    doi:10.1609/aaai.v39i8.32849 , urldate =

    Wang, Siyu and Chen, Cailian and Le, Xinyi and Xu, Qimin and Xu, Lei and Zhang, Yanzhou and Yang, Jie , year = 2025, month = apr, journal =. doi:10.1609/aaai.v39i8.32849 , urldate =. arXiv , langid =:2412.19663 , primaryclass =

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

  11. [11]

    and Alam, Md Ferdous and Nobari, Amin Heyrani and Ahmed, Faez , year = 2025, month = may, number =

    Doris, Anna C. and Alam, Md Ferdous and Nobari, Amin Heyrani and Ahmed, Faez , year = 2025, month = may, number =. doi:10.48550/arXiv.2505.14646 , howpublished =. arXiv , langid =:2505.14646 , primaryclass =

    work page doi:10.48550/arxiv.2505.14646 2025

  12. [12]

    Ocker, Felix and Menzel, Stefan and Sadik, Ahmed and Rios, Thiago , year = 2025, month = mar, number =. From. doi:10.48550/arXiv.2503.04417 , urldate =. arXiv , langid =:2503.04417 , primaryclass =

    work page doi:10.48550/arxiv.2503.04417 2025

  13. [13]

    and Altun, O

    Herrmann, K. and Altun, O. and Wolniak, P. and Mozgova, I. and Lachmayer, R. , year = 2021, doi =. Methodischer Aufbau von Entwicklungsumgebungen Nach Dem Generative Parametric Design Approach , booktitle =

    work page 2021

  14. [14]

    Agentic Large Language Models, a Survey , volume=

    Plaat, Aske and Van Duijn, Max and Van Stein, Niki and Preuss, Mike and Van der Putten, Peter and Batenburg, Kees Joost , year=. Agentic Large Language Models, a Survey , volume=. doi:10.1613/jair.1.18675 , journal=

    work page doi:10.1613/jair.1.18675

  15. [15]

    Understanding the planning of

    Huang, Xiaocheng , year =. Understanding the planning of. arXiv preprint , eprint =

    work page

  16. [16]

    arXiv preprint , eprint =

    Yao, Shunyu and Zhao, Jeffrey and Yu, Dian and Du, Nan and Shafran, Izhak and Narasimhan, Karthik and Cao, Yuan , year =. arXiv preprint , eprint =

    work page

  17. [17]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Toolformer: Language Models Can Teach Themselves to Use Tools , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  18. [18]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Reflexion: Language Agents with Verbal Reinforcement Learning , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

    work page

  19. [19]

    arXiv preprint , eprint =

    Shen, Zhuocheng , year =. arXiv preprint , eprint =

    work page

  20. [20]

    Computer-Aided Design and Applications , volume =

    Advanced Computer Aided Design Methods for Integrated Virtual Product Development Processes , author =. Computer-Aided Design and Applications , volume =

    work page

  21. [21]

    2020 , edition =

    Structural Engineering Handbook , editor =. 2020 , edition =

    work page 2020

  22. [22]

    Shimada, Kenji , year = 2006, month = jan, journal =. Current

    work page 2006

  23. [23]

    Wu, Rundi and Xiao, Chang and Zheng, Changxi , date =. 2021. 2021 , pages =. doi:10.1109/ICCV48922.2021.00670 , url =

    work page doi:10.1109/iccv48922.2021.00670 2021

  24. [24]

    and Willis, Karl D

    Jayaraman, Pradeep Kumar and Sanghi, Aditya and Lambourne, Joseph G. and Willis, Karl D. D. and Davies, Thomas and Shayani, Hooman and Morris, Nigel , year =. arXiv , langid =:2006.10211 , primaryclass =

    work page arXiv 2006

  25. [25]

    2011 , journal =

    Advanced Computer Aided Design Methods for Integrated Virtual Product Development Processes , author =. 2011 , journal =

    work page 2011

  26. [26]

    Principles and Practices of

    Sharma, Vikram and Sharma, Vikrant and Shukla, Om Ji , year =. Principles and Practices of. doi:10.1201/9781003350842 , isbn =

    work page doi:10.1201/9781003350842

  27. [27]

    IEEE Software 40(4), 30–38 (2023) https://doi.org/10.1109/MS.2023.3265877

    Ebert, Christof and Louridas, Panos and Ebert, Christof , year =. Generative. IEEE Software , volume =. doi:10.1109/MS.2023.3265877 , issue_date =

    work page doi:10.1109/ms.2023.3265877 2023

  28. [28]

    Generative

    Brynjolfsson, Erik and Li, Danielle and Raymond, Lindsey R , date =. Generative. doi:10.3386/w31161 , url =

    work page doi:10.3386/w31161

  29. [29]

    Challenges and Opportunities in the Integration of Generative AI with Computer-Aided Design , journal=

    Berger, Elias and Dammann, Maximilian Peter and Mehlst. Challenges and Opportunities in the Integration of Generative AI with Computer-Aided Design , journal=. 2025 , publisher =. doi:10.1017/pds.2025.10102 , pages=

    work page doi:10.1017/pds.2025.10102 2025

  30. [30]

    Xu, Xiang and Willis, Karl D. D. and Lambourne, Joseph G. and Cheng, Chin-Yi and Jayaraman, Pradeep Kumar and Furukawa, Yasutaka , date =. 2022 , eprint =

    work page 2022

  31. [31]

    and Willis, Karl D

    Xu, Xiang and Jayaraman, Pradeep Kumar and Lambourne, Joseph G. and Willis, Karl D. D. and Furukawa, Yasutaka , date =. Hierarchical. 2023 , eprint =

    work page 2023

  32. [32]

    Sequence to Sequence Learning with Neural Networks

    Sutskever, Ilya and Vinyals, Oriol and Le, Quoc V. , year =. Sequence to. doi:10.48550/arXiv.1409.3215 , urldate =. arXiv , keywords =:1409.3215 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1409.3215

  33. [33]

    Attention Is All You Need

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N. and Kaiser, Lukasz and Polosukhin, Illia , date =. Attention. doi:10.48550/arXiv.1706.03762 , url =. 1706.03762 , eprinttype =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1706.03762

  34. [34]

    and Contero, Manuel and Company, Pedro , date =

    Camba, Jorge D. and Contero, Manuel and Company, Pedro , date =. Parametric. 2016 , journaltitle =. doi:10.1016/j.cad.2016.01.003 , url =

    work page doi:10.1016/j.cad.2016.01.003 2016

  35. [35]

    Balancing

    Zhou, Yuxuan and Keuper, Margret and Fritz, Mario , date =. Balancing. doi:10.48550/arXiv.2408.13586 , url =. 2408.13586 , eprinttype =

    work page doi:10.48550/arxiv.2408.13586

  36. [36]

    Goodfellow, Ian J. and. Generative. 2014 , month = jun, number =. doi:10.48550/arXiv.1406.2661 , urldate =. arXiv , keywords =:1406.2661 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1406.2661 2014

  37. [37]

    Willis, Karl D. D. and Pu, Yewen and Luo, Jieliang and Chu, Hang and Du, Tao and Lambourne, Joseph G. and. Fusion 360. 2021 , month = may, number =. doi:10.48550/arXiv.2010.02392 , urldate =. arXiv , keywords =:2010.02392 , primaryclass =

    work page doi:10.48550/arxiv.2010.02392 2021

  38. [38]

    Text2CAD: Generating Sequential

    Khan, Mohammad Sadil and Sinha, Sankalp and Sheikh, Talha Uddin and Stricker, Didier and Ali, Sk Aziz and Afzal, Muhammad Zeshan , date =. 2024 , eprint =. doi:10.48550/arXiv.2409.17106 , url =

    work page doi:10.48550/arxiv.2409.17106 2024

  39. [39]

    Wu, Yonghui and Schuster, Mike and Chen, Zhifeng and Le, Quoc V. and Norouzi, Mohammad and Macherey, Wolfgang and Krikun, Maxim and Cao, Yuan and Gao, Qin and Macherey, Klaus and Klingner, Jeff and Shah, Apurva and Johnson, Melvin and Liu, Xiaobing and Kaiser, Lukasz and Gouws, Stephan and Kato, Yoshikiyo and Kudo, Taku and Kazawa, Hideto and Stevens, Kei...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1609.08144

  40. [40]

    Adam: A Method for Stochastic Optimization

    Kingma, Diederik P. and Ba, Jimmy , year =. Adam:. doi:10.48550/arXiv.1412.6980 , urldate =. arXiv , keywords =:1412.6980 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1412.6980

  41. [41]

    Nash, Charlie and Ganin, Yaroslav and Eslami, S. M. Ali and Battaglia, Peter W. , year =. doi:10.48550/arXiv.2002.10880 , urldate =. arXiv , keywords =:2002.10880 , primaryclass =

    work page doi:10.48550/arxiv.2002.10880 2002

  42. [42]

    doi:10.48550/arXiv.2007.11301 , urldate =

    Carlier, Alexandre and Danelljan, Martin and Alahi, Alexandre and Timofte, Radu , year =. doi:10.48550/arXiv.2007.11301 , urldate =. arXiv , keywords =:2007.11301 , primaryclass =

    work page doi:10.48550/arxiv.2007.11301 2007

  43. [43]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina , year =. doi:10.48550/arXiv.1810.04805 , urldate =. arXiv , keywords =:1810.04805 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1810.04805

  44. [44]

    Journal of Computational Design and Engineering , volume =

    Zhang, Shuming and Guan, Zhidong and Jiang, Hao and Ning, Tao and Wang, Xiaodong and Tan, Pingan , year =. Journal of Computational Design and Engineering , volume =

    work page

  45. [45]

    doi:10.48550/ARXIV.2106.02711 , shorttitle =

    Para, Wamiq Reyaz and Bhat, Shariq Farooq and Guerrero, Paul and Kelly, Tom and Mitra, Niloy and Guibas, Leonidas and Wonka, Peter , year =. doi:10.48550/arXiv.2106.02711 , urldate =. arXiv , keywords =:2106.02711 , primaryclass =

    work page doi:10.48550/arxiv.2106.02711

  46. [46]

    2019 , publisher =

    Advanced CAD Modeling: Explicit, Parametric, Free-Form CAD and Re-engineering , author =. 2019 , publisher =. doi:https://doi.org/10.1007/978-3-030-02399-7 , series =

    work page doi:10.1007/978-3-030-02399-7 2019

  47. [47]

    Language Models Are Unsupervised Multitask Learners , author =

    work page

  48. [48]

    Proceedings of the 30th International Conference on Machine Learning, ICML 2013 , pages=

    Rectifier Nonlinearities Improve Neural Network Acoustic Models , author=. Proceedings of the 30th International Conference on Machine Learning, ICML 2013 , pages=. 2013 , publisher=

    work page 2013

  49. [49]

    Advances in Neural Information Processing Systems , volume=

    Layer normalization , author=. Advances in Neural Information Processing Systems , volume=. 2016 , publisher=

    work page 2016

  50. [50]

    arXiv , howpublished =:2208.10555 , primaryclass =

    Dupont, Elona and Cherenkova, Kseniya and Kacem, Anis and Ali, Sk Aziz and Arzhannikov, Ilya and Gusev, Gleb and Aouada, Djamila , year =. arXiv , howpublished =:2208.10555 , primaryclass =

    work page arXiv

  51. [51]

    2022 , eprint=

    Scaling Language Models: Methods, Analysis and Insights from Training Gopher , author=. 2022 , eprint=

    work page 2022

  52. [52]

    CAD-SIGNet: CAD Language Inference from Point Clouds Using Layer-Wise Sketch Instance Guided Attention , year=

    Khan, Mohammad Sadil and Dupont, Elona and Ali, Sk Aziz and Cherenkova, Kseniya and Kacem, Anis and Aouada, Djamila , booktitle=. CAD-SIGNet: CAD Language Inference from Point Clouds Using Layer-Wise Sketch Instance Guided Attention , year=

    work page

  53. [53]

    Future prospects of computer-aided design (CAD) – A review from the perspective of artificial intelligence (AI), extended reality, and 3D printing , journal =

    Bonsa. Future prospects of computer-aided design (CAD) – A review from the perspective of artificial intelligence (AI), extended reality, and 3D printing , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.rineng.2022.100478 , url =

    work page doi:10.1016/j.rineng.2022.100478 2022

  54. [54]

    Proceedings of the 31st International Conference on Neural Information Processing Systems , pages =

    van den Oord, Aaron and Vinyals, Oriol and Kavukcuoglu, Koray , title =. Proceedings of the 31st International Conference on Neural Information Processing Systems , pages =. 2017 , isbn =

    work page 2017

  55. [55]

    2017 , url=

    Ilya Loshchilov and Frank Hutter , booktitle=. 2017 , url=

    work page 2017

  56. [56]

    CAD-Assistant: Tool-augmented VLLMs as generic CAD task solvers.arXiv preprint arXiv:2412.13810, 2024

    Mallis, Dimitrios and Karadeniz, Ahmet Serdar and Cavada, Sebastian and Rukhovich, Danila and Foteinopoulou, Niki and Cherenkova, Kseniya and Kacem, Anis and Aouada, Djamila , date =. 2025 , eprint =. doi:10.48550/arXiv.2412.13810 , url =

    work page doi:10.48550/arxiv.2412.13810 2025

  57. [57]

    Penev and Bryan Weissinger and M

    Adam Urbanczyk and Jeremy Wright and thebluedirt and Marcus Boyd and Lorenz and Innovations Technology Solutions and Hasan Yavuz Özderya and Bruno Agostini and Jojain and Michael Greminger and Seth Fischer and Justin Buchanan and cactrot and huskier and Ruben and Iulian Onofrei and Miguel Sánchez de León Peque and Martin Budden and Hecatron and Peter Boin...

    work page doi:10.5281/zenodo.10513848

  58. [58]

    GenCAD: Image- Conditioned Computer-Aided Design Generation with Transformer-Based Contrastive Representation and Diffu- sion Priors

    Alam, Md Ferdous and Ahmed, Faez , date =. http://arxiv.org/abs/2409.16294 , urldate =. 2025 , eprint =. doi:10.48550/arXiv.2409.16294 , url =

    work page doi:10.48550/arxiv.2409.16294 2025

  59. [59]

    Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL) , pages=

    Prefix-Tuning: Optimizing Continuous Prompts for Generation , author=. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics (ACL) , pages=. 2021 , url=

    work page 2021

  60. [60]

    LoRA: Low-Rank Adaptation of Large Language Models

    Hu, Edward J. and Shen, Yelong and Wallis, Phillip and. 2021 , month = oct, number =. doi:10.48550/arXiv.2106.09685 , urldate =. arXiv , langid =:2106.09685 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2106.09685 2021

  61. [61]

    2024 , eprint=

    StarCoder 2 and The Stack v2: The Next Generation , author=. 2024 , eprint=

    work page 2024

  62. [62]

    A Point Set Generation Network for 3D Object Reconstruction from a Single Image

    Fan, Haoqiang and Su, Hao and Guibas, Leonidas J. , year =. A Point Set Generation Network for. CoRR , volume =. 1612.00603 , archiveprefix =

    work page internal anchor Pith review Pith/arXiv arXiv

  63. [63]

    CodeBLEU: a Method for Automatic Evaluation of Code Synthesis

    Shuo Ren and Daya Guo and Shuai Lu and Long Zhou and Shujie Liu and Duyu Tang and Neel Sundaresan and Ming Zhou and Ambrosio Blanco and Shuai Ma , title =. CoRR , volume =. 2020 , url =. 2009.10297 , timestamp =

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  64. [64]

    Qwen2.5-Coder Technical Report

    Qwen2. 5-Coder Technical Report , author=. arXiv preprint arXiv:2409.12186 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  65. [65]

    2024 , eprint=

    CodeGemma: Open Code Models Based on Gemma , author=. 2024 , eprint=

    work page 2024

  66. [66]

    DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence

    DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence , author =. arXiv preprint arXiv:2401.14196 , year =. 2401.14196 , archivePrefix =

    work page internal anchor Pith review Pith/arXiv arXiv

  67. [67]

    Small Language Models are the Future of Agentic AI

    Small Language Models are the Future of Agentic AI , author =. 2025 , eprint =. doi:10.48550/arXiv.2506.02153 , howpublished =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2506.02153 2025

  68. [68]

    2024 , eprint=

    Tuning Language Models by Proxy , author=. 2024 , eprint=

    work page 2024

  69. [69]

    AgentNet: A scalable framework for multi-step agent trajectory generation.arXiv preprint arXiv:2501.00000,

    Task-Adaptive CAD Generation via Decoder-Only Pretrained Transformer , author=. arXiv preprint arXiv:2501.00000 , year=

    work page arXiv

  70. [70]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Mamba-CAD: State Space Model for 3D Computer-Aided Design Generative Modeling , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2025 , month=. doi:10.1609/aaai.v39i5.32531 , abstractNote=

    work page doi:10.1609/aaai.v39i5.32531 2025

  71. [71]

    2024 , howpublished=

    StableLM 3B 4E1T , author=. 2024 , howpublished=

    work page 2024

  72. [72]

    2024 , eprint=

    Gemma 2: Improving Open Language Models at a Practical Size , author=. 2024 , eprint=

    work page 2024

  73. [73]

    2020 , eprint=

    Visual Transformers: Token-based Image Representation and Processing for Computer Vision , author=. 2020 , eprint=

    work page 2020

  74. [74]

    The Annals of Mathematical Statistics , volume=

    On a test of whether one of two random variables is stochastically larger than the other , author=. The Annals of Mathematical Statistics , volume=. 1947 , publisher=

    work page 1947

  75. [75]

    Special Issue: CARs & FOF '92 , volume =

    Functional Features for Design in Mechanical Engineering , author =. Special Issue: CARs & FOF '92 , volume =. doi:10.1016/0166-3615(93)90111-D , abstract =

    work page doi:10.1016/0166-3615(93)90111-d

  76. [76]

    Journal of the Royal Statistical Society , volume=

    On the interpretation of ^2 from contingency tables, and the calculation of P , author=. Journal of the Royal Statistical Society , volume=. 1922 , publisher=

    work page 1922

  77. [77]

    Topology Optimization:

    Bends. Topology Optimization:

    work page

  78. [78]

    doi:10.1016/j.cad.2025.103926 , keywords =

    Lv, Chaofan and Bao, Jinsong , year = 2025, journal =. doi:10.1016/j.cad.2025.103926 , keywords =

    work page doi:10.1016/j.cad.2025.103926 2025

  79. [79]

    IEEE Transactions on Industrial Informatics , volume =

    work page

  80. [80]

    doi:10.48550/arXiv.2507.09792 , urldate =

    Govindarajan, Prashant and Baldelli, Davide and Pathak, Jay and Fournier, Quentin and Chandar, Sarath , year = 2026, month = jan, number =. doi:10.48550/arXiv.2507.09792 , urldate =. arXiv , langid =:2507.09792 , primaryclass =

    work page doi:10.48550/arxiv.2507.09792 2026

Showing first 80 references.