Pith. sign in

REVIEW 4 major objections 6 minor 83 references

Toward Knowledge-Guided AI for Inverse Design in Manufacturing: A Perspective on Domain, Physics, and Human-AI Synergy

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This perspective argues that combining expert-guided problem definition, physics-informed machine learning, and LLM-based natural-language interfaces makes AI-driven inverse design in manufacturing practical in sparse-data…

desk verdict A sensible, well-organized perspective that honestly marks its own limits, but the central integration claim remains an untested proposal. read the letter →

arxiv 2506.00056 v2 pith:LK25LSKQ submitted 2025-05-29 cs.AI physics.comp-ph

Hugon Lee , Hyeonbin Moon , Junhyeong Lee , Seunghwa RYu This is my paper
classification cs.AIphysics.comp-ph
keywords inversedesignphysics-informedmachinelearninglargelanguagemodelshuman-AIcollaborationinjectionmoldingsmartmanufacturingdomainknowledgeautomation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Purely data-driven AI inverse design often fails in real manufacturing because datasets are sparse and biased, design spaces are large, and process constraints are stringent. This perspective argues that combining three complementary pillars—expert-guided problem definition, physics-informed machine learning, and large-language-model-based interfaces—can overcome those failures and improve data efficiency, generalization, and interpretability. The argument is made through a review of each pillar and a concrete injection-molding workflow, not through new experiments. The paper's thesis is that inverse design should be a knowledge-guided collaboration between humans and AI rather than a black-box optimization loop.

What carries the argument

The central object is the proposed three-pillar framework, summarized in the inverse-design workflow of the injection molding case study. The load-bearing mechanisms are the specific ways each pillar intervenes: expert-guided variable reduction and constraint embedding shrink the feasible space; physics-informed neural networks or operator models add PDE residuals computed by automatic differentiation to the loss function; and large language models are grounded through retrieval-augmented generation and tool or multi-agent orchestration so they act as interfaces rather than standalone predictors. The framework's claim is that these mechanisms are complementary: physics covers the data-scarce regime, LLM interfaces cover the interaction regime, and expert definition makes the other two tractable.

What would settle it

Take the injection-molding workflow of the paper and run it on a fixed sparse, biased dataset, first with the full three-pillar pipeline and then with the same surrogate model trained without the physics-loss term and without expert variable pruning; if the knowledge-guided version does not yield higher validated design quality or better held-out accuracy, the claimed improvements in data efficiency and generalization do not hold. A second check is to shift the operating environment and see whether pruned variables that were ranked irrelevant under the original data regain influence; if they do, the problem-definition pillar has removed the wrong variables.

Watch

Extended reading notes

Core claim

The central claim, stated in the abstract and introduction, is that integrating expert-guided problem definition, physics-informed machine learning, and LLM-based natural-language interfaces can improve data efficiency, generalization, and interpretability for AI-driven inverse design in manufacturing. On the paper's terms, no single pillar is sufficient: domain knowledge makes objectives and constraints physically meaningful, physics-informed machine learning keeps forward models faithful under sparse or biased data, and LLM interfaces make the workflow accessible and auditable. The injection molding example shows all three operating together—experts prune variables and set machine, material, and geometry constraints, physics-informed neural networks embed heat transfer or rheology into the loss, and a retrieval-augmented multi-agent LLM supports interactive process design. If the paper is right, this integrated approach, rather than any one algorithm, is what makes inverse design practical in industry.

Load-bearing premise

The framework's benefit depends on the process's governing physics being known well enough to encode in a model, and on experts being able to prune design variables without dropping ones that become influential under other conditions; the paper concedes that many manufacturing equations are only partially known and that aggressive variable reduction carries that risk.

Editorial extensions

If this is right

  • Inverse design can move from interpolation-friendly dense-data settings to the extrapolation settings common in manufacturing, where training data cover only a subregion of the feasible space.
  • Embedding manufacturing constraints during problem definition should reduce infeasible candidate designs and yield solutions that respect machine limits, material windows, and geometry restrictions.
  • Physics-informed forward models should deliver physically consistent predictions with fewer samples, lowering the experimental or simulation budget needed for optimization.
  • LLM-based interfaces with retrieval augmentation should let non-specialists query, steer, and question design workflows while keeping human accountability clear.
  • Optimized designs from any of these pipelines still require validation by experiment or high-fidelity simulation before deployment, and LLM outputs need safeguards against hallucination.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension the paper leaves implicit is to benchmark the full three-pillar pipeline against a plain data-driven surrogate on the same biased injection-molding dataset, isolating each pillar's contribution by ablation; the paper predicts the integrated version recovers better validated optima.
  • The common industrial case may be hybrid modeling with partially known physics rather than full governing equations; the paper flags this limitation but does not develop systematic hybrid benchmarks, so those benchmarks are a natural next step.
  • If LLM interfaces succeed, the economic value may come less from prediction accuracy than from reduced expert time in problem reformulation and result interpretation, shifting the bottleneck from modeling to interface trust.
  • Because environmental variables shift optimal settings, a robust system will need to treat process optima as conditional on context; the paper mentions contextual variables in injection molding but leaves robustness across environment shifts as an open problem.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript is a perspective proposing that AI-driven inverse design in manufacturing should be built as an integrated framework with three pillars: expert-guided problem definition, physics-informed machine learning (PIML), and LLM-based natural-language interfaces. It reviews each pillar in turn, illustrates the combination through an injection-molding case study, and closes with open challenges. The central assertion, stated in the Abstract and Section 1, is that this integration 'can improve data efficiency, generalization, and interpretability' relative to purely data-driven approaches.

Significance. If the proposed framework is validated, it would provide a useful synthesis and research agenda for an active area. The manuscript is clearly organized, covers a broad literature, and explicitly acknowledges limitations within each pillar, which is a strength for a perspective. Its concrete injection-molding illustration helps ground the discussion. However, the paper is a perspective rather than an empirical study: no integrated system is implemented or measured, and the central claim is a hypothesis. The value of the paper currently lies in framing, not in demonstrated results; this is acceptable for the genre only if the claim is carefully reframed as a research program and the composition of pillars is subjected to critical analysis.

major comments (4)
  1. [Abstract and Section 1] The central claim that integrating the three pillars 'can improve data efficiency, generalization, and interpretability' is not supported by any integrated demonstration. Section 6, especially Figure 6d, is a schematic workflow that chains together separately published works (Refs 81–83); there is no end-to-end implementation or quantitative comparison against a purely data-driven baseline. Please either reframe the claim explicitly as a testable research hypothesis/agenda, or add a proof-of-concept or benchmark plan with concrete metrics for data efficiency, generalization, and interpretability.
  2. [Sections 3 and 4] The manuscript discusses limitations of each pillar separately but does not analyze how the pillars interact. For example, expert-guided variable pruning that removes variables that 'could become influential under different operating conditions' (Section 3) may also remove boundary conditions or coupling terms needed for the PDE residuals used in PIML (Section 4); similarly, LLM-mediated problem definition (Section 5) introduces hallucination risk that could corrupt the very constraints the first pillar is meant to embed. Please add a section that analyzes at least two plausible failure modes of the integrated pipeline and proposes safeguards or design rules to mitigate them.
  3. [Section 4] The framework's second pillar depends on governing equations being sufficiently known and embeddable, yet Section 4 concedes that many manufacturing processes have equations that are 'partially known, empirically derived, or too complex to embed directly.' This concession narrows the domain in which the integrated claim is expected to hold. Please state explicitly the applicability conditions under which the three-pillar integration is expected to help, and when it is not, and connect that boundary to the injection-molding illustration in Section 6.
  4. [Sections 3, 6, and reference list] Many of the concrete examples used to illustrate the framework, particularly in injection molding, are the authors' own prior publications (e.g., Refs 7, 10–13, 15, 31, 80, 83). This is not inappropriate for a perspective, but it means the supporting evidence is partly self-referential and may not convince a general reader. Please balance the illustration with independent external examples, or explicitly note where the authors' own line of work is being used as a case study rather than as independent validation.
minor comments (6)
  1. [Section 3] The word 'formultion' in 'establish feasible variable ranges during problem formultion' should be corrected to 'formulation'.
  2. [Acknowledgements] 'grand funded' should be 'grant funded', and the final sentence contains a duplicated period ('(N10250154)..').
  3. [Section 2] 'can effectively utilized with many inverse design/optimization algorithms' is ungrammatical; change to 'can be effectively utilized in many inverse design/optimization algorithms.'
  4. [Section 4] The sentence 'Physics-informed variants, such as physics-informed DeepONet (PIDON), incorporate PDE residuals' is fine, but the later sentence 'Physics-informed DeepONet (PIDON) incorporate PDE residuals' has subject-verb disagreement; unify the grammar.
  5. [Throughout] Hyphenation and spacing of compound terms such as 'LLM -based', 'multi -agent', and 'physics -informed' are inconsistent; use a single style (e.g., 'LLM-based', 'multi-agent', 'physics-informed').
  6. [Figure 6d] The workflow arrows in Figure 6d are dense; consider labeling the five steps more prominently and indicating which components are existing implementations (Refs 81–83) versus proposed extensions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a perspective with no fitted predictions or derivations, and its self-citations are illustrative pointers rather than load-bearing justifications.

full rationale

This is a perspective paper, not a derivation chain. It proposes an integrated framework combining expert-guided problem definition, physics-informed machine learning, and LLM-based interfaces, and illustrates the proposal with injection molding. The paper fits no parameters, derives no equations, and makes no quantitative prediction that could be shown equivalent to its inputs by construction. Self-citations (e.g., Refs. 7, 10-13, 15, 80, 83) are used as examples of prior applications or as pointers to fuller reviews, not as the sole justification for the central 'can improve' claim. No uniqueness theorem, fitted input renamed as prediction, or ansatz smuggled in via citation is present. The paper's own caveats—'aggressive reduction carries the risk of excluding variables that, while seemingly minor in the available dataset, could become influential under different operating conditions' (Section 3), 'Many manufacturing problems involve governing equations that are partially known, empirically derived, or too complex to embed directly' (Section 4), and 'hallucination remains a persistent risk' (Section 5)—are limitations on the strength and completeness of the proposal. They weaken the evidence for the integrated claim, but they do not make any step definitionally circular. The main weakness is that the composition of the three pillars is asserted rather than demonstrated end-to-end; that is an evidence or correctness concern, not circularity.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No free parameters or invented entities. Three domain assumptions are load-bearing for the proposal; all are acknowledged as challenges in the paper itself.

assumptions (3)
  • domain assumption Domain expertise can reliably prune design variables and define feasible operating ranges before data acquisition.
    The proposed framework assumes expert-guided problem definition reduces dimensionality without excluding influential variables; Section 3 explicitly acknowledges the risk of over-pruning but still adopts the benefit.
  • domain assumption The governing physics of manufacturing processes is known well enough to be embedded in PIML architectures.
    Section 4 states many manufacturing problems have partially known, empirically derived, or too complex equations, which undermines the pillar's applicability; the claim depends on PIML being usable.
  • domain assumption LLM-based interfaces can be made reliable, secure, and accountable in high-stakes manufacturing contexts.
    Section 5 acknowledges hallucination, data privacy, and accountability risks but still asserts LLMs extend accessibility; the claim depends on these being manageable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Toward Knowledge-Guided AI for Inverse Design in Manufacturing: A Perspective on Domain, Physics, and Human-AI Synergy." pith.science (2026). https://pith.science/paper/LK25LSKQ

@misc{pith2026250600056,
  author       = {Pith},
  title        = {Pith review of: Toward Knowledge-Guided AI for Inverse Design in Manufacturing: A Perspective on Domain, Physics, and Human-AI Synergy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LK25LSKQ}},
  note         = {Machine review of arXiv:2506.00056}
}
read the original abstract

Artificial intelligence (AI) is reshaping inverse design in manufacturing, enabling high-performance discovery in materials, products, and processes. However, purely data-driven approaches often struggle in realistic manufacturing settings characterized by sparse data, high-dimensional design spaces, and complex constraints. This perspective proposes an integrated framework built on three complementary pillars: domain knowledge to establish physically meaningful objectives and constraints while removing variables with limited relevance, physics-informed machine learning to enhance generalization under limited or biased data, and large language model-based interfaces to support intuitive, human-centered interaction. Using injection molding as an illustrative example, we demonstrate how these components can operate in practice and conclude by highlighting key challenges for applying such approaches in realistic manufacturing environments.

Figures

Figures reproduced from arXiv: 2506.00056 by the authors.

Figure 1
Figure 1. Schematic workflow of AI-based inverse design process, highlighting the stages where AI contributes. Concepts discussed in this Perspective—expert-guided problem definition (Section 3), PIML (Section 4), and LLM-based interfaces (Section 5)—are marked within the workflow [PITH_FULL_IMAGE:figures/full_fig_p028_1.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

83 extracted references · 61 canonical work pages

  1. [1]

    A., Xu, Y

    Micchelli, C. A., Xu, Y. & Zhang, H. Universal Kernels. Journal of Machine Learning Research vol. 7 (2006)

    2006

  2. [2]

    & White, H

    Hornik, K., Stinchcombe, M. & White, H. Multilayer Feedforward Networks Are Universal Approximators. Neural networks vol. 2 359–366 (Elsevier, 1989)

    1989

  3. [3]

    & Kim, J

    Kang, Y. & Kim, J. ChatMOF: An Artificial Intelligence System for Predicting and Generating Metal -Organic Frameworks Using Large Language Models . Nature communications vol. 15 4705 (Nature Publishing Group UK London, 2024)

    2024

  4. [4]

    Lee, I. et al. Uncovering the Relationship between Metal Elements and Mechanical Stability for Metal–Organic Frameworks. ACS Appl. Mater. Interfaces 16, 52162–52178 (2024)

    work page 2024

  5. [5]

    & Grossman, J

    Xie, T. & Grossman, J. C. Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties. Phys. Rev. Lett. 120, 145301 (2018)

    2018

  6. [6]

    Serles, P. et al. Ultrahigh Specific Strength by Bayesian Optimization of Carbon Nanolattices. Adv. Mater. 37, 2410651 (2025)

    work page 2025

  7. [7]

    Lee, H. et al. Bayesian optimization of tailgate rib structures enhancing structural stiffness under manufacturing constraints of injection molding. J. Manuf. Process. 134, 739 –748 (2025)

    work page 2025

  8. [8]

    & Kang, N

    Oh, S., Jung, Y., Kim, S., Lee, I. & Kang, N. Deep Generative Design: Integration of Topology Optimization and Generative Models. J. Mech. Des. 141, 111405 (2019)

    work page 2019

Show all 83 references
  1. [9]

    & Ryu, S

    Cho, H., Lee, H. & Ryu, S. Sequential multi -objective Bayesian optimization of air-cooled battery thermal management system with spoiler integration. J. Energy Storage 114, 115586 (2025)

    2025

  2. [10]

    & Ryu, S

    Jung, J., Park, K., Lee, H., Cho, B. & Ryu, S. Comparative Study of Multi -objective Bayesian Optimization and NSGA-III based Approaches for Injection Molding Process. Adv. Theory Simul. 7, 2400135 (2024)

    2024

  3. [11]

    & Ryu, S

    Lee, H., Lee, M., Jung, J., Lee, I. & Ryu, S. Enhancing Injection Molding Optimization for SFRPs Through Multi-Fidelity Data-Driven Approaches Incorporating Prior Information in Limited Data Environments. Adv. Theory Simul. 7, 2400130 (2024)

    2024

  4. [12]

    -Y., Kim, H., Nam, K., Kang, D

    Kim, J. -Y., Kim, H., Nam, K., Kang, D. & Ryu, S. Development of an injection molding production condition inference system based on diffusion model. J. Manuf. Syst. 79, 162–178 (2025)

    2025

  5. [13]

    & Ryu, S

    Jung, J., Park, K., Cho, B., Park, J. & Ryu, S. Optimization of injection molding process using multi -objective bayesian optimization and constrained generative inverse design networks. J. Intell. Manuf. 34, 3623–3636 (2023)

    2023

  6. [14]

    Brunton, S. L. & Kutz, J. N. Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control . (Cambridge University Press, Cambridge, 2019). doi:10.1017/9781108380690

    2019 doi

  7. [15]

    Lee, J. et al. Machine learning -based inverse design methods considering data characteristics and design space size in materials design and manufacturing: a review. Mater. Horiz. 10, 5436–5456 (2023)

    2023

  8. [16]

    & Ryu, S

    Yu, J. & Ryu, S. Deep Reinforcement Learning -Based Optimization for the Shape of an Adhesive Pillar With Enhanced Adhesion Strength. J. Mech. Des. 1–12 (2025) doi:10.1115/1.4068747

    2025 doi

  9. [17]

    Makatura, L. et al. How Can Large Language Models Help Humans in Design and Manufacturing? Preprint at https://doi.org/10.48550/arXiv.2307.14377 (2023)

    2023 doi arXiv

  10. [18]

    Li, Y. et al. Large Language Models for Manufacturing. Preprint at https://doi.org/10.48550/arXiv.2410.21418 (2024)

    2024 doi arXiv

  11. [19]

    & Meierhofer, J

    Wulf, J. & Meierhofer, J. The Impact of Large Language Models on Task Automation in Manufacturing Services. Procedia CIRP 134, 1089–1094 (2025)

    2025

  12. [20]

    Karniadakis, G. E. et al. Physics-informed machine learning. Nat. Rev. Phys. 3, 422– 440 (2021)

    2021

  13. [21]

    Hao, Z. et al. Physics-Informed Machine Learning: A Survey on Problems, Methods and Applications. Preprint at https://doi.org/10.48550/arXiv.2211.08064 (2023)

    2023 doi arXiv

  14. [22]

    Han, X.-Q. et al. Ai-driven inverse design of materials: Past, present, and future. Chin. Phys. Lett. 42, 027403 (2025)

    2025

  15. [23]

    J., Rzonca, A., Lakhnori, R

    Plathottam, S. J., Rzonca, A., Lakhnori, R. & Iloeje, C. O. A review of artificial intelligence applications in manufacturing operations. J. Adv. Manuf. Process. 5, e10159 (2023)

    2023

  16. [24]

    F., Chang, Q., Gao, R

    Arinez, J. F., Chang, Q., Gao, R. X., Xu, C. & Zhang, J. Artificial Intelligence in Advanced Manufacturing: Current Status and Future Outlook. J. Manuf. Sci. Eng. 142, 110804 (2020)

    2020

  17. [25]

    X., Krüger, J., Merklein, M., Möhring, H

    Gao, R. X., Krüger, J., Merklein, M., Möhring, H. -C. & Váncza, J. Artificial Intelligence in manufacturing: State of the art, perspectives, and future directions. CIRP Ann. 73, 723–749 (2024)

    2024

  18. [26]

    Mainini, L. et al. Analytical Benchmark Problems for Multifidelity Optimization Methods. Preprint at https://doi.org/10.48550/arXiv.2204.07867 (2022)

    2022 doi arXiv

  19. [27]

    & Rasmussen, C

    Williams, C. & Rasmussen, C. Gaussian Processes for Regression. in Advances in Neural Information Processing Systems vol. 8 (MIT Press, 1995)

    1995

  20. [28]

    & Ryu, S

    Jo, J., Park, K., Song, H., Lee, H. & Ryu, S. Innovative 3D printing of mechanoluminescent composites: Vat photopolymerization meets machine learning. Addit. Manuf. 90, 104324 (2024)

    2024

  21. [29]

    Kashinath, K. et al. Physics-informed machine learning: case studies for weather and climate modelling. Philos. Trans. R. Soc. Math. Phys. Eng. Sci. 379, 20200093 (2021)

    2021

  22. [30]

    & Karniadakis, G

    Raissi, M., Yazdani, A. & Karniadakis, G. E. Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations. Science 367, 1026–1030 (2020)

    2020

  23. [31]

    Moon, H. et al. Physics-informed neural network -based discovery of hyperelastic constitutive models from extremely scarce data. Comput. Methods Appl. Mech. Eng. 446, 118258 (2025)

    2025

  24. [32]

    & Juanes, R

    Haghighat, E., Raissi, M., Moure, A., Gomez, H. & Juanes, R. A Physics-Informed Deep Learning Framework for Inversion and Surrogate Modeling in Solid Mechanics . Computer Methods in Applied Mechanics and Engineering vol. 379 113741 (Elsevier, 2021)

    2021

  25. [33]

    & Karniadakis, G

    Raissi, M., Perdikaris, P. & Karniadakis, G. E. Physics -informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J. Comput. Phys. 378, 686–707 (2019)

    2019

  26. [34]

    G., Pearlmutter, B

    Baydin, A. G., Pearlmutter, B. A., Radul, A. A. & Siskind, J. M. Automatic differentiation in machine learning: a survey. Preprint at https://doi.org/10.48550/arXiv.1502.05767 (2018)

    2018 doi arXiv

  27. [35]

    & Mahoney, M

    Krishnapriyan, A., Gholami, A., Zhe, S., Kirby, R. & Mahoney, M. W. Characterizing possible failure modes in physics-informed neural networks. Adv. Neural Inf. Process. Syst. 34, 26548–26560 (2021)

    2021

  28. [36]

    & Karniadakis, G

    Lu, L., Jin, P., Pang, G., Zhang, Z. & Karniadakis, G. E. Learning Nonlinear Operators via DeepONet Based on the Universal Approximation Theorem of Operators . Nature machine intelligence vol. 3 218–229 (Nature Publishing Group UK London, 2021)

    2021

  29. [37]

    Kovachki, N. et al. Neural operator: Learning maps between function spaces with applications to pdes. J. Mach. Learn. Res. 24, 1–97 (2023)

    2023

  30. [38]

    & Karniadakis, G

    Goswami, S., Bora, A., Yu, Y. & Karniadakis, G. E. Physics -Informed Deep Neural Operator Networks. Preprint at https://doi.org/10.48550/arXiv.2207.05748 (2022)

    2022 doi arXiv

  31. [39]

    & Perdikaris, P

    Wang, S., Wang, H. & Perdikaris, P. Learning the solution operator of parametric partial differential equations with physics -informed DeepONets. Sci. Adv. 7, eabi8605 (2021)

    2021

  32. [40]

    Li, Z., Kovachki, N. B. & Azizzadenesheli, K. Burigede liu, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Fourier neural operator for parametric partial differential equations. in International Conference on Learning Representations vol. 2 4 (2021)

    2021

  33. [41]

    Li, Z. et al. Physics-Informed Neural Operator for Learning Partial Differential Equations. ACM IMS J. Data Sci. 1, 1–27 (2024)

    2024

  34. [42]

    Go, M.-S., Lim, J. H. & Lee, S. Physics-Informed Neural Network-Based Surrogate Model for a Virtual Thermal Sensor with Real -Time Simulation. International Journal of Heat and Mass Transfer vol. 214 124392 (Elsevier, 2023)

    2023

  35. [43]

    S., Reynvaan, J

    Sabathiel, S., Sanchis-Alepuz, H., Wilson, A. S., Reynvaan, J. & Stipsitz, M. Neural Network-Based Reconstruction of Steady -State Temperature Systems with Unknown Material Composition . Scientific Reports vol. 14 22265 (Nature Publishing Group UK London, 2024)

    2024

  36. [44]

    Jin, H. et al. Characterization and Inverse Design of Stochastic Mechanical Metamaterials Using Neural Operators. Adv. Mater. 2420063 (2025) doi:10.1002/adma.202420063

    2025 doi

  37. [45]

    & Wang, F

    Yang, S., Peng, S., Guo, J. & Wang, F. A review on physics -informed machine learning for monitoring metal additive manufacturing process. Adv Manuf 1, 1–28 (2024)

    2024

  38. [46]

    Pooch, U. W. & Wall, J. A. Discrete Event Simulation: A Practical Approach . (CRC Press, Boca Raton, 2024). doi:10.1201/9781003575054

    2024 doi

  39. [47]

    & Uddin, Z

    Ganga, S. & Uddin, Z. Exploring Physics -Informed Neural Networks: From Fundamentals to Applications in Complex Systems. Preprint at https://doi.org/10.48550/arXiv.2410.00422 (2024)

    2024 doi arXiv

  40. [48]

    Kasilingam, S. et al. Physics-based and data-driven hybrid modeling in manufacturing: a review. Prod. Manuf. Res. 12, 2305358 (2024)

    2024

  41. [49]

    Wang, J., Li, Y., Gao, R. X. & Zhang, F. Hybrid physics-based and data-driven models for smart manufacturing: Modelling, simulation, and explainability. J. Manuf. Syst. 63, 381– 391 (2022)

    2022

  42. [50]

    Li, J. et al. CAD-Llama: Leveraging Large Language Models for Computer -Aided Design Parametric 3D Model Generation. Preprint at https://doi.org/10.48550/arXiv.2505.04481 (2025)

    2025 doi arXiv

  43. [51]

    C., Alam, M

    Doris, A. C., Alam, M. F., Nobari, A. H. & Ahmed, F. CAD-Coder: An Open-Source Vision-Language Model for Computer -Aided Design Code Generation. Preprint at https://doi.org/10.48550/arXiv.2505.14646 (2025)

    2025 doi arXiv

  44. [52]

    M., Man, B

    Edwards, K. M., Man, B. & Ahmed, F. Sketch2Prototype: rapid conceptual design exploration and prototyping with generative AI. Proc. Des. Soc. 4, 1989–1998 (2024)

    2024

  45. [53]

    & Farimani, A

    Jadhav, Y. & Farimani, A. B. Large Language Model Agent as a Mechanical Designer. Preprint at https://doi.org/10.48550/arXiv.2404.17525 (2025)

    2025 doi arXiv

  46. [54]

    Karray, F., Alemzadeh, M., Saleh, J. A. & Arab, M. N. Human-Computer Interaction: Overview on State of the Art. Int. J. Smart Sens. Intell. Syst. 1, 137–159 (2008)

    2008

  47. [55]

    Designs that fit

    Te’Eni, D. Designs that fit. Hum.-Comput. Interact. Manag. Inf. Syst. Found. 205–224 (2015)

    2015

  48. [56]

    Designing the User Interface: Strategies for Effective Human - Computer Interaction

    Shneiderman, B. Designing the User Interface: Strategies for Effective Human - Computer Interaction. (Pearson Education India, 2010)

    2010

  49. [57]

    Usability Engineering, Morgan Kaufmann Publishers Inc

    Nielsen, J. Usability Engineering, Morgan Kaufmann Publishers Inc. Burlingt. Mass. San Franc. Calif. 2008 (1994)

    1994

  50. [58]

    & Zaraki, A

    Ghamati, K., Banitalebi Dehkordi, M. & Zaraki, A. Towards AI-Powered Applications: The Development of a Personalised LLM for HRI and HCI. Sensors 25, 2024 (2025)

    2025

  51. [59]

    & Liu, A

    Wang, X., Jiang, Z., Xiong, Y. & Liu, A. Human -LLM collaboration in generative design for customization. J. Manuf. Syst. 80, 425–435 (2025)

    2025

  52. [60]

    Xu, Q. et al. Generative AI and DT integrated intelligent process planning: a conceptual framework. Int. J. Adv. Manuf. Technol. 133, 2461–2485 (2024)

    2024

  53. [61]

    Freire, S. K. et al. Knowledge Sharing in Manufacturing using Large Language Models: User Evaluation and Model Benchmarking. Preprint at https://doi.org/10.48550/arXiv.2401.05200 (2024)

    2024 doi arXiv

  54. [62]

    & Farimani, A

    Chandrasekhar, A., Chan, J., Ogoke, F., Ajenifujah, O. & Farimani, A. B. AMGPT: A Large Language Model for Contextual Querying in Additive Manufacturing . Additive Manufacturing Letters vol. 11 100232 (Elsevier, 2024)

    2024

  55. [63]

    Fan, H., Liu, X., Fuh, J. Y. H., Lu, W. F. & Li, B. Embodied intelligence in manufacturing: leveraging large language models for autonomous industrial robotics. J. Intell. Manuf. 36, 1141–1157 (2025)

    2025

  56. [64]

    Fan, H. et al. AutoMEX: Streamlining Material Extrusion with AI Agents Powered by Large Language Models and Knowledge Graphs . Materials & Design 113644 (Elsevier, 2025)

    2025

  57. [65]

    & Weyrich, M

    Xia, Y., Shenoy, M., Jazdi, N. & Weyrich, M. Towards autonomous system: flexible modular production system enhanced with large language model agents. in 2023 IEEE 28th International Conference on Emerging Technologies and Factory Automation (ETFA) 1–8 (IEEE, 2023)

    2023

  58. [66]

    & Mercangöz, M

    Vyas, J. & Mercangöz, M. Autonomous Industrial Control using an Agentic Framework with Large Language Models. Preprint at https://doi.org/10.48550/arXiv.2411.05904 (2024)

    2024 doi

  59. [67]

    & Kovalenko, I

    Lim, J., Vogel-Heuser, B. & Kovalenko, I. Large language model-enabled multi-agent manufacturing systems. in 2024 IEEE 20th International Conference on Automation Science and Engineering (CASE) 3940–3946 (IEEE, 2024)

    2024

  60. [68]

    & Farimani, A

    Jadhav, Y., Pak, P. & Farimani, A. B. LLM -3D Print: Large Language Models To Monitor and Control 3D Printing. Preprint at https://doi.org/10.48550/arXiv.2408.14307 (2025)

    2025 doi

  61. [69]

    Jeon, J. et al. ChatCNC: Conversational Machine Monitoring via Large Language Model and Real -Time Data Retrieval Augmented Generation . Journal of Manufacturing Systems vol. 79 504–514 (Elsevier, 2025)

    2025

  62. [70]

    Yuksekgonul, M. et al. Optimizing generative AI by backpropagating language model feedback. Nature 639, 609–616 (2025)

    2025

  63. [71]

    Opsahl-Ong, K. et al. Optimizing Instructions and Demonstrations for Multi -Stage Language Model Programs. Preprint at https://doi.org/10.48550/arXiv.2406.11695 (2024)

    2024 doi arXiv

  64. [72]

    Zhang, J. et al. AFlow: Automating Agentic Workflow Generation. Preprint at https://doi.org/10.48550/arXiv.2410.10762 (2025)

    2025 doi arXiv

  65. [73]

    Zhou, H. et al. Multi-Agent Design: Optimizing Agents with Better Prompts and Topologies. Preprint at https://doi.org/10.48550/arXiv.2502.02533 (2025)

    2025 doi

  66. [74]

    Wu, S. et al. Optimas: Optimizing Compound AI Systems with Globally Aligned Local Rewards. Preprint at https://doi.org/10.48550/arXiv.2507.03041 (2025)

    2025 doi

  67. [75]

    Kapania, S., Wang, R., Li, T. J. -J., Li, T. & Shen, H. ‘I’m Categorizing LLM as a Productivity Tool’: Examining Ethics of LLM Use in HCI Research Practices. Proc ACM Hum-Comput Interact 9, CSCW102:1-CSCW102:26 (2025)

    2025

  68. [76]

    Liu, X. et al. Large Language Models and Causal Inference in Collaboration: A Comprehensive Survey. in Findings of the Association for Computational Linguistics: NAACL 2025 (eds Chiruzzo, L., Ritter, A. & Wang, L.) 7668 –7684 (Association for Computational Linguistics, Albuque...

    2025 doi

  69. [77]

    Carlini, N. et al. Extracting Training Data from Large Language Models. in 2633 – 2650 (2021)

    2021

  70. [78]

    It’s a Fair Game

    Zhang, Z. et al. “It’s a Fair Game”, or Is It? Examining How Users Navigate Disclosure Risks and Benefits When Using LLM-Based Conversational Agents. in Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems 1–26 (Association for Computing Machinery, New ...

    2024

  71. [79]

    Zheng, R., Tanner, R. I. & Fan, X. -J. Injection Molding: Integration of Theory and Modeling Methods. (Springer Science & Business Media, 2011)

    2011

  72. [80]

    -Y., Yu, J., Kim, H

    Kim, J. -Y., Yu, J., Kim, H. & Ryu, S. DRL -Based Injection Molding Process Parameter Optimization for Adaptive and Profitable Production. Preprint at https://doi.org/10.48550/arXiv.2505.10988 (2025)

    2025 doi arXiv

  73. [81]

    & Lucchetta, G

    Pieressa, A., Baruffa, G., Sorgato, M. & Lucchetta, G. Enhancing weld line visibility prediction in injection molding using physics -informed neural networks. J. Intell. Manuf. (2024) doi:10.1007/s10845-024-02460-w

    2024 doi

  74. [82]

    Physics-Informed Neural Networks for Inverse Modeling of Viscosity in Injection Molding

    Lindqvist, E. Physics-Informed Neural Networks for Inverse Modeling of Viscosity in Injection Molding. Masters Theses Math. Sci. (2025)

    2025

  75. [83]

    -Y., Kim, H., Lee, I

    Lee, J., Kim, J. -Y., Kim, H., Lee, I. & Ryu, S. IM -Chat: A Multi-agent LLM-based Framework for Knowledge Transfer in Injection Molding Industry. Preprint at https://doi.org/10.48550/arXiv.2507.15268 (2025). Figures and tables Figure 1. Schematic workflow of AI -based inverse...

    2025 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.