Pith. sign in

REVIEW 4 major objections 4 minor 93 references

DASyR-LLM: Domain-Aware Symbolic Regression with LLMs for Kinetic Model Discovery

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Embedding an LLM that critiques and proposes rate laws into symbolic regression reduces the experiments needed to recover kinetic models by 41.7 to 79.3 percent.

desk verdict Solid, honest integration of an LLM critic-generator into an SR-MBDoE loop for kinetics; the headline speedup is plausible but not yet cleanly measured because the manual parsing step can leak the target structure. read the letter →

arxiv 2608.05120 v1 pith:Y6REVGAQ submitted 2026-08-05 cs.LG cs.CEcs.SC

Roberto Aliaga Medina , Paulina Quintanilla , Antonio del Rio Chanona This is my paper · ORCID
classification cs.LGcs.CEcs.SC
keywords symbolicregressionkineticmodeldiscoverylargelanguagemodelsmodel-baseddesignofexperimentsratelawidentificationdomainknowledgeinjectioniterativeexperimentbioprocessmodelling
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

This paper tries to establish that a large language model can act as a domain-knowledge injection point inside an iterative symbolic-regression pipeline for discovering reaction rate laws from noisy concentration data. At each iteration, the LLM critiques the top symbolic-regression candidates for physicochemical plausibility and proposes new rate expressions, which then compete with the regression candidates under the same information-criterion-based selection and model-based design of experiments. Across four synthetic kinetics case studies, the guided loop recovers the ground-truth rate law in 41.7–79.3% fewer iterations than regression alone, with the LLM directly proposing the correct structure in more than half of the guided runs and validation accuracy ($R^2>0.98$) unchanged. If true, each saved iteration is one saved wet-lab experiment, so the contribution is a practical reduction in experimental effort rather than a new statistical estimator.

What carries the argument

The central object is the two-step LLM module inserted into the iterative discovery loop at the point where the best symbolic-regression rate candidates have been parameter-estimated and scored. In step one the LLM receives the global reaction, species names, the top five candidate rate laws with their NLL and AIC values, and a few-shot example, and returns a free-text mechanistic critique. In step two, that critique is fed forward with instructions to propose three structurally distinct, mechanistically explainable rate laws with symbolic parameters; the proposals are parsed manually into the same parameter-estimation and AIC-ranking machinery as the regression candidates, and only the two best models then drive the model-based design of the next experiment. The mechanism works by expanding the candidate pool with knowledge-guided structures while keeping acceptance statistical, so the LLM biases the search without overriding the data.

What would settle it

Replace the four textbook-derived ground truths with chemically plausible rate laws that are not published anywhere the model could have memorized (for example, a rational expression with an unusual inhibition term), run the same guided loop, and check whether the 41.7–79.3% iteration reduction persists; if it collapses, the savings come from memorized retrieval rather than domain reasoning. A second check: have the LLM proposals parsed by an automated, blinding script and compare how often ground-truth-like structures survive parsing versus random rewordings.

Watch

Extended reading notes

Core claim

Embedding an LLM module between parameter estimation and model selection inside a symbolic-regression plus model-based-design-of-experiments loop reduces the number of iterations needed to identify the ground-truth kinetic model (or a near-equivalent expression) by 41.7–79.3% relative to the same loop without the LLM, while producing models with equivalent out-of-sample accuracy ($R^2>0.98$ across all case studies). The LLM performs two functions at each iteration: it gives a qualitative physicochemical critique of the five best regression candidates, using the symbolic forms, fitted parameter magnitudes, and AIC values, and it proposes three new structurally distinct rate expressions informed by that critique and its embedded chemical knowledge. These proposals are parsed manually, fitted by ODE integration, and then ranked against the regression candidates with the same AIC criterion, so the LLM does not bypass the statistical selection pipeline. In 9 of 16 guided runs the final model was proposed directly by the LLM; in the remaining runs it steered the search toward functional forms that symbolic regression subsequently recovered.

Load-bearing premise

The speedup is measured against ground truths the LLM may have seen during pretraining, so the load-bearing assumption is that the LLM's proposals come from physicochemical reasoning over the presented regression candidates rather than memorized retrieval of textbook rate laws, and that the manual parsing of free-text proposals does not preferentially preserve ground-truth-like structures.

Editorial extensions

If this is right

  • Every iteration saved in the discovery loop is a model-design-of-experiments experiment not run, so the 41.7–79.3% iteration reduction translates directly into fewer wet-lab experiments needed to reach a validated kinetic model.
  • The LLM acts as an active generator, not just a filter: in 9 of 16 guided runs it proposed the final model structure directly, and in the other 7 it steered symbolic regression toward the correct functional form.
  • Guidance does not trade away accuracy: validation $R^2$ stayed above 0.98 for both algorithms, so the efficiency gain is not paid for in predictive performance.
  • The symbolic-regression component remains load-bearing: when it was removed, an LLM without access to raw data failed to converge in one of two runs and needed six iterations in the other, indicating the two components are complementary.
  • A smaller backbone LLM retained most of the benefit on the bioprocess case, converging in 2–5 iterations versus 1–3 for the larger model and 3–10 for baseline regression, suggesting the approach does not depend on a frontier-scale model.

Reading between the lines

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

  • If the speedup generalizes, the economic value of the loop scales with per-experiment cost; in process development where a single batch experiment costs days of bench time, a 40–80% cut in iteration count is likely worth more than the added compute per iteration.
  • A direct probe of the memorization concern would be to run the same loop on chemically plausible rate laws that are absent from the LLM's pretraining corpus; stable iteration savings there would separate physicochemical reasoning from retrieval.
  • The with-data ablation hints at an untested variant: an LLM that reads raw concentration traces can propose the correct structure at iteration one, suggesting a spectrum from regression-anchored guidance to fully LLM-driven discovery depending on how much data the LLM is allowed to see.
  • The framework's benefit concentrates in case studies where baseline symbolic regression struggles (nitrous oxide decomposition and the bioprocess), so the expected gain in a new application is likely largest when mechanism-rich rational or inhibited rate laws are plausible and smallest for near-power-law kinetics.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper introduces DASyR-LLM, a framework that embeds an LLM module into the ADoK-S iterative symbolic-regression and model-based design-of-experiments loop. At each iteration the LLM critiques the top SR candidates on physicochemical grounds and proposes new rate expressions, which are then parameter-estimated and ranked by AIC together with SR candidates. The framework is evaluated on four in silico kinetic case studies (toluene hydrodealkylation, nitrous oxide decomposition, a theoretical isomerisation, and a bacterial protein production bioprocess) with four independent runs per algorithm per case study. The reported central result is a 41.7–79.3% reduction in iterations to recover the ground-truth or near-equivalent kinetic model relative to SR alone, with equivalent validation R² above 0.98. Ablation studies remove SR and replace the 14B LLM with an 8B model, and the paper publicly releases code and full LLM logs.

Significance. If the central claim holds, the framework offers a practical route to cutting wet-lab experimental effort in kinetic model discovery, and it is a useful contribution to the growing literature on LLM-guided scientific discovery. The paper is unusually transparent in several respects: the source code and complete LLM input/output logs are archived, the synthetic-data generation and validation protocols are described in detail, parameter confidence intervals are reported, and the authors explicitly acknowledge the pretraining-exposure and manual-parsing limitations. Those strengths make the main claims testable. However, the headline speedup is currently co-produced with human parsing and is supported by only four runs per condition, so the quantitative claim is not yet established as a clean property of the framework.

major comments (4)
  1. [Section 3.1.4 and Section 7] The manual, non-blinded parsing of LLM proposals is load-bearing for the central speedup claim. Section 3.1.4 states that LLM-proposed expressions are 'parsed manually into a structured format' and the authors opted against structured output, while Section 7 lists this parsing as a current limitation. Since the authors know the four ground-truth rate laws in Section 4, a human parser can unconsciously regularize ambiguous free-text proposals toward target-like structures or discard equally defensible alternatives. The 9/16 runs flagged as 'directly suggested by the LLM' in Table 3 and the reported 41.7–79.3% iteration reduction both depend on this conversion. This is testable with the archived logs: re-run the pipeline with a blind parser or with automated structured-output parsing and compare iteration counts, and report inter-parser agreement on the existing logs.
  2. [Section 5.1, Table 3, and Figure 5] The headline reduction is based on only four independent runs per case-study/algorithm cell, and no statistical significance test or confidence interval is reported for the iteration reductions. The bioprocess baseline, for example, spans 3–10 iterations while the LLM-guided runs span 1–3, so the mean reduction is sensitive to a single slow baseline run; the isomerisation case shows substantial run-to-run overlap. I recommend reporting per-case bootstrap confidence intervals for the reduction and a paired or permutation test across runs, or otherwise increasing the number of replicates before claiming that the reduction is consistent.
  3. [Section 5.1, Table 3, runs 11–12] The iteration-reduction metric conflates speed-to-convergence with success rate. Two baseline nitrous-oxide runs reach the 12-iteration budget without recovering the ground-truth model and are counted as 12 iterations in the comparison. This makes the reported percentage reduction partly a measure of avoiding failed runs rather than a pure measure of faster identification of the target. Because the central claim is 'iterations required to identify the ground-truth model', I recommend reporting censored time-to-convergence and success rates separately, or a combined metric that distinguishes non-convergence from convergence at the budget cap.
  4. [Section 6.3] The pretraining-memorization risk is acknowledged but not resolved, and it directly affects the interpretation of the 9/16 'LLM directly proposed' outcomes. The authors correctly note that the prompt does not provide case identity, but the reaction and species names in the prompt may be sufficient for a model pretrained on kinetics textbooks to retrieve the Langmuir–Hinshelwood, inhibition, and Monod forms used as benchmarks. The bioprocess ablation partially mitigates this concern because the system is adapted and the LLM alone is unreliable, but it is a single case study. A concrete test would be to add case studies with mechanistically valid rate laws that are unlikely to appear verbatim in the pretraining corpus, or to run a memorization probe on held-out textbook rate laws, and to report whether the speedup persists.
minor comments (4)
  1. [Table 3 vs. Table S38 (run 26)] There is a data inconsistency in the bacterial production case: Table 3 reports the run-26 model as B·S/((0.277·B+2.088)·(S+1.823)), while Table S38 lists the fitted parameter k3 as 1.7436, not 1.823. Please correct one of the two values.
  2. [Supplementary Information S6.2] The prompt text contains a typo: 'if you p2ropose a term' should read 'if you propose a term'.
  3. [Abstract and Section 5.1] The abstract's '41.7–79.3%' range is a mean reduction across case studies, but the abstract does not state that it is a mean. Adding 'mean reduction' would make the summary more precise, given the large run-to-run variability shown in Figure 4.
  4. [Section 3.4 and Figure 7] The computational-cost comparison is clear, but the wall-clock measurements include LLM inference run on a remote HPC cluster while other components run on a local workstation; reporting this as a single 'per-iteration cost' may be sensitive to network and queue variability. A brief statement on whether LLM times include network latency would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the key iteration-reduction results are empirical measurements on external synthetic benchmarks, and neither the LLM proposals nor the fitted parameters are defined in terms of the target models.

full rationale

DASyR-LLM's central claim is that adding an LLM critique/proposal module to ADoK-S reduces the iterations needed to recover ground-truth kinetic models. The comparison is self-contained: both algorithms run under identical experimental budgets on synthetic data generated from external literature rate laws (Fogler; Levenspiel; Marin et al.; Forster et al.), and the reported iteration counts, AIC rankings, and validation R2 values are measured outcomes, not quantities fitted into the comparison. LLM-proposed expressions are parsed manually (Sec. 3.1.4) and then subjected to the same ODE-integration parameter estimation and AIC selection as SR candidates; the ground-truth structure is not used to define the score function, and no fitted parameter is renamed as a prediction. The self-citations to ADoK-S and PI-ADoK are contextual and not load-bearing for the headline reduction, and the paper invokes no uniqueness theorem. The acknowledged limitations—manual parsing and possible LLM pretraining exposure (Secs. 6.3, 7)—are contamination/robustness concerns, not circular derivations: they do not make the iteration metric equal to its inputs by construction. Hence no circular step can be exhibited under the required standard.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The framework itself postulates no new physical quantities. The ledger records the algorithm settings and domain assumptions that the benchmark comparison relies on. The main soft spot is the unverified assumption that the LLM's successful proposals come from general reasoning rather than from pretraining exposure to the textbook case studies, together with hand-chosen hyperparameters that are not swept.

free parameters (2)
  • Experiment budget and LLM module settings = n_B=12, k_SR=5, n_LLM=3, n_gen=3
    Hand-chosen hyperparameters that define how many SR candidates are critiqued, how many LLM proposals are made, and how many iterations are allowed; no sensitivity sweep is reported for these settings.
  • Synthetic data noise and sampling constants = sigma=0.1, n_p=30, two initial experiments
    Data-generation choices that define benchmark difficulty; these are assumptions about experimental conditions rather than fitted parameters, but they shape all iteration counts.
assumptions (6)
  • domain assumption Ground-truth rate laws and parameters from the cited literature are the correct targets for the synthetic data.
    Section 4 defines each case study by a literature rate law; the entire iteration-count comparison depends on these targets being representative and correctly implemented.
  • domain assumption All species are measurable and measurement noise is zero-mean Gaussian with known variance sigma=0.1.
    Section 3.1.1 generates data under this assumption; Section 7 acknowledges real wet-lab data may have missing observations and systematic bias.
  • standard math AIC with independently estimated per-species variances is a valid model-selection criterion for ranking kinetic models.
    Section 3.1.5 derives AIC from NLL; the framework's stopping and MBDoE selection depend on AIC ranking.
  • domain assumption The multistart L-BFGS-B parameter estimator reaches sufficiently good optima that AIC rankings are trustworthy.
    Section 3.1.3 uses 20 random starts and notes global optimality is not guaranteed; poor initial guesses may discard structurally correct LLM candidates (Section 5.2).
  • ad hoc to paper The LLM's proposals reflect domain reasoning rather than memorized retrieval of the three textbook case studies.
    Section 6.3 states pretraining exposure cannot be determined; the LLM-only ablation is used as evidence but cannot exclude partial leakage of literature-derived structures.
  • domain assumption PySR with the given operator set and budgets is an adequate SR oracle and fair baseline.
    Section 3.1.2 configures PySR; the baseline ADoK-S uses the same backbone, so the comparison isolates the LLM module.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DASyR-LLM: Domain-Aware Symbolic Regression with LLMs for Kinetic Model Discovery." pith.science (2026). https://pith.science/paper/Y6REVGAQ

@misc{pith2026260805120,
  author       = {Pith},
  title        = {Pith review of: DASyR-LLM: Domain-Aware Symbolic Regression with LLMs for Kinetic Model Discovery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y6REVGAQ}},
  note         = {Machine review of arXiv:2608.05120}
}
abstract

Kinetic model discovery is a central challenge in chemical engineering, as accurate rate expressions are essential for understanding and controlling chemical and biological processes. Symbolic regression (SR) has emerged as a powerful data-driven approach for identifying interpretable kinetic models, but usually operates without domain knowledge, often exploring physicochemically implausible models. Large language models (LLMs) offer a promising avenue for injecting domain expertise into this search. Here, we introduce an LLM-guided SR framework, embedding an LLM module within an iterative SR algorithm for automated kinetic model discovery. The LLM performs two roles at each iteration: (1) a qualitative physicochemical critique of the best SR candidates, and (2) the proposal of new candidate rate expressions guided by the SR-generated models and embedded chemical knowledge. Our framework is evaluated on four in silico case studies of increasing complexity, spanning heterogeneous catalysis and bioprocess systems. Results show the LLM-guided framework reduces iterations to identify the ground-truth model by $41.7-79.3\%$ versus a state-of-the-art SR framework, with the LLM directly proposing the correct model structure in over half of the guided runs. In practical settings, where each iteration typically requires a new wet-lab experiment, this translates into a substantial reduction in experimental effort. Predictive performance on an independent validation set is equivalent between both approaches, with $R^2>0.98$ in all case studies. Ablation studies indicate that both the SR component and the LLM scale contribute to this performance, with a reduced-size LLM largely retaining discovery efficiency. These findings demonstrate that LLMs can effectively inject domain knowledge into scientific model discovery, paving the way toward fully automated, domain-aware kinetic modelling pipelines.

Figures

Figures reproduced from arXiv: 2608.05120 by the authors.

Figure 1
Figure 1. Schematic of the LLM-guided SR workflow (DASyR-LLM), comprising nine steps. The LLM module (step [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the two-stage SR workflow using data from a single experiment (theoretical isomerisation case [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Structure of the LLM module used for model discovery guidance. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Number of iterations until termination (convergence or exhaustion of the experimental budget) for all runs, [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Mean iteration reduction (%) achieved by the LLM-guided SR framework relative to the baseline SR, [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]
Figure 6
Figure 6. Figure 6: Wall-clock time versus number of iterations until termination for all 32 experimental runs. Each point [PITH_FULL_IMAGE:figures/full_fig_p019_6.png]
Figure 7
Figure 7. Figure 7: Mean time per iteration broken down by component (SR, LLM inference, parameter estimation, and MBDoE), [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: compares the predictive performance of the final models, measured by R2 on the independent validation dataset described in Section 3.3, which spans interior, boundary, and extrapolation regions of the input space. Across all case studies and species, both algorithms ac…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

93 extracted references · 71 canonical work pages

  1. [1]

    Rasmuson, B

    A. Rasmuson, B. Andersson, L. Olsson and R. Andersson, inIntroduction, Cambridge University Press, 2014, p. 1–9

    2014

  2. [2]

    M. Á. de Carvalho Servia, I. O. Sandoval, K. K. M. Hii, K. Hellgardt, D. Zhang and E. Antonio del Rio Chanona, Digital Discovery, 2024,3, 954–968

    2024

  3. [3]

    X. Huan, J. Jagalur and Y . Marzouk,Acta Numerica, 2024,33, 715–840

    2024

  4. [4]

    N. S. Eyke, W. H. Green and K. F. Jensen,Reaction Chemistry & Engineering, 2020,5, 1963–1972

    2020

  5. [5]

    Cranmer,Interpretable Machine Learning for Science with PySR and SymbolicRegression.jl, arXiv, 2023, preprint, arXiv:2305.01582, 2023, https://doi.org/10.48550/arXiv.2305.01582

    M. Cranmer,Interpretable Machine Learning for Science with PySR and SymbolicRegression.jl, arXiv, 2023, preprint, arXiv:2305.01582, 2023, https://doi.org/10.48550/arXiv.2305.01582

    arXiv doi:10.48550/arxiv.2305.01582 2023

  6. [6]

    A. W. Rogers, A. Lane, C. Mendoza, S. Watson, A. Kowalski, P. Martin and D. Zhang,Chemical Engineering Science, 2024,300, 120580

    2024

  7. [7]

    Makke and S

    N. Makke and S. Chawla,Artificial Intelligence Review, 2024,57, 2

    2024

  8. [8]

    de Vries, S

    S. de Vries, S. W. Keemink and M. A. J. van Gerven,Symbolic Discovery of Stochastic Differential Equations with Genetic Programming, arXiv, 2026, preprint, arXiv:2603.09597, https://doi.org/10.48550/arXiv.2603.09597

    work page doi:10.48550/arxiv.2603.09597 2026

Show all 93 references
  1. [9]

    Y . A. Radwan, G. Kronberger and S. Winkler, Computer Aided Systems Theory – EUROCAST 2024, Cham, 2025, pp. 157–171

    2024

  2. [10]

    M. Á. de Carvalho Servia, K. K. M. . Hii, K. Hellgardt, D. Zhang and E. A. del Rio Chanona,Chem. Sci., 2026, 17, 7533–7550

    2026

  3. [11]

    Forster, D

    T. Forster, D. Vázquez, M. N. Cruz-Bournazou, A. Butté and G. Guillén-Gosálbez,Computers & Chemical Engineering, 2023,170, 108108

    2023

  4. [13]

    Z. T. Wilson and N. V . Sahinidis,Computers & Chemical Engineering, 2017,106, 785–795. 26 LLM-guided Symbolic Regression for Kinetic Model Discovery

    2017

  5. [14]

    Prabhu, N

    S. Prabhu, N. Kosir, M. V . Kothare and S. Rangarajan,Industrial & Engineering Chemistry Research, 2025,64, 2601–2615

    2025

  6. [15]

    S. Han, U. Awasthi and G. M. Bollas,Journal of Manufacturing Systems, 2025,80, 734–748

    2025

  7. [16]

    L. S. Keren, A. Liberzon and T. Lazebnik,Scientific Reports, 2023,13, 1249

    2023

  8. [17]

    P. A. K. Reinbold, L. M. Kageorge, M. F. Schatz and R. O. Grigoriev,Nature Communications, 2021,12, 3219

    2021

  9. [18]

    Y . Zhu, S. Chen and X. Li,Canadian Geotechnical Journal, 2025,62, 1–14

    2025

  10. [19]

    Eichhorn, A

    S. Eichhorn, A. Mohapatra and C. Goebel, Proceedings of the 16th ACM International Conference on Future and Sustainable Energy Systems, New York, NY , USA, 2025, p. 92–107

    2025

  11. [20]

    M. Á. de Carvalho Servia, A. Ali, I. Orson Sandoval, K. K. Hii, K. Hellgardt, D. Zhang, M. Mehmet and A. del Rio Chanona,Physics Informed Symbolic Regression for Data-Efficient Kinetic Model Discovery, SSRN, 2026, preprint, SSRN:6709721, http://dx.doi.org/10.2139/ssrn.6709721

    2026 doi

  12. [21]

    Dong and J

    J. Dong and J. Zhong,ACM Comput. Surv., 2025,57, 1–37

    2025

  13. [22]

    Merler, K

    M. Merler, K. Haitsiukevich, N. Dainese and P. Marttinen, Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (V olume 4: Student Research Workshop), 2024, p. 589–606

    2024

  14. [23]

    Z. Guo, S. Wang, Y . Tian, J. Yang, H. Yu, X. Na, L. Kovács, L. Li, P. A. Ioannou and F.-Y . Wang,Proceedings of the National Academy of Sciences, 2025,122, e2516995122

    2025

  15. [24]

    Shojaee, N.-H

    P. Shojaee, N.-H. Nguyen, K. Meidani, A. B. Farimani, K. D. Doan and C. K. Reddy, Forty-second International Conference on Machine Learning, 2025

    2025

  16. [25]

    A. Z. Bideh and J. Gryak,LLM-ODE: Data-driven Discovery of Dynamical Systems with Large Language Models, arXiv, 2026, preprint, arXiv:2603.20910, https://doi.org/10.48550/arXiv.2603.20910

    2026 doi arXiv

  17. [26]

    Zhang, Q

    H. Zhang, Q. Chen, B. Xue, W. Banzhaf and M. Zhang,LLM-Meta-SR: In-Context Learning for Evolving Selection Operators in Symbolic Regression, arXiv, 2026, preprint, arXiv:2505.18602, https://doi.org/10.48550/arXiv.2505. 18602

    2026 doi

  18. [27]

    Pourcel, C

    J. Pourcel, C. Colas and P.-Y . Oudeyer,Self-Improving Language Models for Evolutionary Program Synthesis: A Case Study on ARC-AGI, arXiv, 2026, preprint, arXiv:2507.14172, https://doi.org/10.48550/arXiv.2507.14172

    2026 doi

  19. [28]

    H. A. Gozeten, X. Zhang, E. Ildiz, E. O. Taga, T. Javidi and S. Oymak,Evolutionary Multi-Task Optimization for LLM-Guided Program Discovery, arXiv, 2026, preprint, arXiv:2605.22613, https://doi.org/10.48550/arXiv.2605. 22613

    2026 doi arXiv

  20. [29]

    Neumann, L

    P. Neumann, L. Cao, D. Russo, V . S. Vassiliadis and A. A. Lapkin,Chemical Engineering Journal, 2020,387, 123412

    2020

  21. [30]

    S. L. Brunton, J. L. Proctor and J. N. Kutz,Proceedings of the National Academy of Sciences, 2016,113, 3932–3937

    2016

  22. [31]

    Lyu and F

    W. Lyu and F. Galvanin,Computers & Chemical Engineering, 2025,202, 109265

    2025

  23. [32]

    G. S. Gusmão, A. P. Retnanto, S. C. da Cunha and A. J. Medford,Catalysis Today, 2023,417, 113701

    2023

  24. [33]

    Sun and R

    W. Sun and R. D. Braatz,Computers & Chemical Engineering, 2020,143, 107103

    2020

  25. [34]

    A. W. Rogers, A. Lane, C. Mendoza, S. Watson, A. Kowalski, P. Martin and D. Zhang,Computers & Chemical Engineering, 2025,197, 109036

    2025

  26. [35]

    Kay and D

    H. Kay and D. Zhang,Journal of Chemical Information and Modeling, 2026, ASAP

    2026

  27. [36]

    Narayanan, M

    H. Narayanan, M. N. Cruz Bournazou, G. Guillén Gosálbez and A. Butté,Chemical Engineering Journal, 2022, 430, 133032

    2022

  28. [37]

    Di Caprio and M

    U. Di Caprio and M. E. Leblebici,Chemical Engineering Science, 2026,321, 122873

    2026

  29. [38]

    S. Jog, D. Vázquez, L. F. Santos, J. A. Caballero and G. Guillén-Gosálbez,Computers & Chemical Engineering, 2024,182, 108563

    2024

  30. [39]

    F. A. R. Lima, M. G. de Moraes, C. M. Rebello, A. G. Barreto, A. R. Secchi, M. B. de Souza and I. B. Nogueira, Chemical Engineering and Processing - Process Intensification, 2025,215, 110350

    2025

  31. [40]

    F. A. R. Lima, M. G. de Moraes, M. E. Leblebici, A. R. Secchi, M. B. de Souza and I. B. Nogueira,Journal of Process Control, 2026,161, 103700

    2026

  32. [41]

    Tabrizi, E

    Z. Tabrizi, E. Barbera, W. Leal da Silva and F. Bezzo,Digital Chemical Engineering, 2025,17, 100276

    2025

  33. [42]

    Rossi, F

    L. Rossi, F. Bezzo and M. Barolo,Computers & Chemical Engineering, 2026,210, 109634. 27 LLM-guided Symbolic Regression for Kinetic Model Discovery

    2026

  34. [43]

    M. Du, Y . Chen, Z. Wang, L. Nie and D. Zhang,LLM4ED: Large Language Models for Automatic Equation Discovery, arXiv, 2024, preprint, arXiv:2405.07761, https://doi.org/10.48550/arXiv.2405.07761

    2024 doi arXiv

  35. [44]

    Shojaee, K

    P. Shojaee, K. Meidani, S. Gupta, A. B. Farimani and C. K. Reddy,LLM-SR: Scientific Equation Discovery via Programming with Large Language Models, arXiv, 2025, preprint, arXiv:2404.18400, https://doi.org/10.48550/ arXiv.2404.18400

    2025 doi arXiv

  36. [45]

    C. Shen, Z. Guo, K. Zuo, W. Huang and M. Sun,LLM-DMD: Large Language Model-based Power System Dynamic Model Discovery, arXiv, 2026, preprint, arXiv:2601.05632, https://doi.org/10.48550/arXiv.2601.05632

    2026 doi

  37. [46]

    Ivanchik, T

    E. Ivanchik, T. Bavshin and A. Hvatov, NeurIPS 2025 AI for Science Workshop, 2025

    2025

  38. [47]

    S. Holt, Z. Qian, T. Liu, J. Weatherall and M. van der Schaar, Advances in Neural Information Processing Systems, 2024, pp. 96325–96366

    2024

  39. [48]

    Spillias, J

    S. Spillias, J. G. D. Rogers, F. Boschetti, E. A. Fulton, M. Guglielmo, S. Y . Yong and R. Trebilco,Methods in Ecology and Evolution, 2026,17, 1303–1321

    2026

  40. [49]

    S. Wahl, R. Schenk, A. Farnoud, J. H. Macke and D. Gedon,A Probabilistic Framework for LLM-Based Model Discovery, arXiv, 2026, preprint, arXiv:2602.18266, https://arxiv.org/abs/2602.18266

    2026 arXiv

  41. [50]

    Taskin, W

    B. Taskin, W. Xie and T. Lazebnik,Scientific Reports, 2026,16, 1614

    2026

  42. [51]

    Grayeli, A

    A. Grayeli, A. Sehgal, O. Costilla-Reyes, M. Cranmer and S. Chaudhuri,Symbolic Regression with a Learned Concept Library, arXiv, 2024, preprint, arXiv:2409.09359, https://doi.org/10.48550/arXiv.2409.09359

    2024 doi arXiv

  43. [52]

    E. S. Saveliev, S. Holt, N. Seedat, D. L. Bentley, J. Weatherall and M. van der Schaar,Influence-Guided Symbolic Regression: Scientific Discovery via LLM-Driven Equation Search with Granular Feedback, arXiv, 2026, preprint, arXiv:2605.29184, https://doi.org/10.48550/arXiv.2605.29184

    2026 doi arXiv

  44. [53]

    Rangarajan,Current Opinion in Chemical Engineering, 2026,52, 101240

    S. Rangarajan,Current Opinion in Chemical Engineering, 2026,52, 101240

    2026

  45. [54]

    R. W. Mee,A Comprehensive Guide to Factorial Two-Level Experimentation, Springer Science & Business Media, 1st edn, 2009

    2009

  46. [55]

    P. A. Brodtkorb,numdifftools, Version 0.9.42, 2025

    2025

  47. [56]

    A. Yang, A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, C. Lv, C. Zheng, D. Liu, F. Zhou, F. Huang, F. Hu, H. Ge, H. Wei, H. Lin, J. Tang, J. Yang, J. Tu, J. Zhang, J. Yang, J. Yang, J. Zhou, J. Zhou, J. Lin, K. Dang, K. Bao, K. Yang, L. Yu, L. Deng, M. L...

    2025 doi arXiv

  48. [57]

    Sutherland, D

    C. Sutherland, D. Hare, P. J. Johnson, D. W. Linden, R. A. Montgomery and E. Droge,Proceedings of the Royal Society B: Biological Sciences, 2023,290, 20231261

    2023

  49. [58]

    J. E. Cavanaugh and A. A. Neath,WIREs Computational Statistics, 2019,11, e1460

    2019

  50. [59]

    W. G. Hunter and A. M. Reiner,Technometrics, 1965,7, 307–323

    1965

  51. [60]

    H. S. Fogler,Elements of Chemical Reaction Engineering, Prentice Hall, Philadelphia, PA, 5th edn, 2016

    2016

  52. [61]

    Levenspiel,Chemical Reaction Engineering, John Wiley & Sons, Nashville, TN, 3rd edn, 1998

    O. Levenspiel,Chemical Reaction Engineering, John Wiley & Sons, Nashville, TN, 3rd edn, 1998

    1998

  53. [62]

    G. B. Marin, G. S. Yablonsky and D. Constales,Kinetics of Chemical Reactions: Decoding Complexity, John Wiley & Sons, 2nd edn, 2019

    2019

  54. [63]

    A. A. Schoepfer, J. Weinreich, R. Laplaza, J. Waser and C. Corminboeuf,Digital Discovery, 2024,3, 2289–2297

    2024

  55. [64]

    K. Nöh, S. Niedenführ, M. Beyß and W. Wiechert,PLoS Computational Biology, 2018,14, e1006533

    2018

  56. [65]

    Bosten, M

    E. Bosten, M. Pardon, K. Chen, V . Koppen, G. V . Herck, M. Hellings and D. Cabooter,Analytical Chemistry, 2024,96, 13699–13709

    2024

  57. [66]

    J. S. Nielsen, M. G. Jacobsen, A. B. Olson, M. P. Sørensen and A. P. Engsig-Karup,Physics-Informed Regression: Parameter Estimation in Parameter-Linear Nonlinear Dynamic Models, arXiv, 2025, preprint, arXiv:2508.19249, https://doi.org/10.48550/arXiv.2508.19249

    2025 doi arXiv

  58. [67]

    Weglarz-Tomczak, J

    E. Weglarz-Tomczak, J. M. Tomczak, A. E. Eiben and S. Brul,Population-based Optimization for Kinetic Parame- ter Identification in Glycolytic Pathway in Saccharomyces cerevisiae, arXiv, 2020, preprint, arXiv:2010.06456, https://doi.org/10.48550/arXiv.2010.06456

    2020 doi arXiv

  59. [68]

    Della Santa,Mathematics, 2024,12, 1201

    F. Della Santa,Mathematics, 2024,12, 1201. 28 LLM-guided Symbolic Regression for Kinetic Model Discovery

    2024

  60. [69]

    Martí, J

    R. Martí, J. A. Lozano, A. Mendiburu and L. Hernando, inMulti-start Methods, ed. R. Martí, P. M. Pardalos and M. G. C. Resende, Springer International Publishing, Cham, 2018, pp. 155–175

    2018

  61. [70]

    Abliz, T

    D. Abliz, T. Artys and G. Ziegmann,Journal of Applied Polymer Science, 2017,134, 45137. 29 LLM-guided Symbolic Regression for Kinetic Model Discovery Supplementary Information 30 LLM-guided Symbolic Regression for Kinetic Model Discovery S1. Performance Metrics Table S1: Perfo...

    2017

  62. [72]

    The current 5 best candidate rate laws found by Symbolic Regression, with their AIC value.,→ * TASK: - Explain why these kinetic equations could be plausible for their global reaction.,→ - Your answer should also include the kinetic equations and their AIC value. - Consider th...

  63. [75]

    best_candidates

    Current best candidate rate laws: {info["best_candidates"]}

  64. [76]

    additional_info

    Additional information: {info["additional_info"]}. * EXAMPLE: BEGINNING OF EXAMPLE

  65. [79]

    Current best candidate rate laws (AIC in parentheses): - (k1 * KCO * CO * KO2 * O2) / (1 + KCO * CO + KO2 * O2)^2 (AIC = 95.2) - (k2 * CO * O2) / (1 + KCO * CO) (AIC = 101.8) - k3 * CO * O2 (AIC = 128.4) - (k4 * CO) / (1 + KCO * CO) (AIC = 110.6) - k5 * CO^2 * O2 (AIC = 140.3)

  66. [80]

    This expression is characteristic of a Langmuir-Hinshelwood mechanism in which both CO and O2 adsorb onto the catalyst surface before reacting

    Additional information: Reaction carried out over a solid metal catalyst at constant temperature.,→ Answer: The rate law r = (k1 * KCO * CO * KO2 * O2) / (1 + KCO * CO + KO2 * O2)^2 (AIC = 95.2) is highly plausible. This expression is characteristic of a Langmuir-Hinshelwood m...

  67. [82]

    The current 5 best candidate rate laws found by Symbolic Regression, with their AIC value and an explanation of why they could be plausible for the global reaction. ,→ ,→ * TASK: - Propose three kinetic equations for the following global reaction, considering the available exp...

  68. [85]

    best_candidates_analysis

    Analysis of current best candidate rate laws: {info["best_candidates_analysis"]}.,→ * EXAMPLE: BEGINNING OF EXAMPLE

  69. [86]

    49 LLM-guided Symbolic Regression for Kinetic Model Discovery

    Global reaction: CO + 1/2 O2→CO2. 49 LLM-guided Symbolic Regression for Kinetic Model Discovery

  70. [88]

    Analysis of current best candidate rate laws: - (k1 * KCO * CO * KO2 * O2) / (1 + KCO * CO + KO2 * O2)^2 (AIC = 95.2) -> Plausible Langmuir-Hinshelwood mechanism with competitive adsorption.,→ - (k2 * CO * O2) / (1 + KCO * CO) (AIC = 101.8)→CO adsorption controls surface cover...

  71. [90]

    Additional information. * TASK: - Propose five kinetic equations for the biomass change in time following global reaction, considering your (bio)chemical & mathematical knowledge.,→ - Ensure all equations are (bio)chemically plausible and structurally consistent with known kin...

  72. [95]

    Carbon dioxide (CO2): CO2

    Products names: ("Carbon dioxide (CO2): CO2"). Reactants names: ("Carbon monoxide (CO): CO", "Oxygen (O2): O2"). Answer: A plausible new kinetic equation is r = (k6 * KCO * CO * (KO2 * O2)^(1/2)) / (1 + KCO * CO + KO2 * O2),→ This expression differs from all previous candidate...

  73. [97]

    Additional information. * TASK: - Propose five kinetic equations for the biomass change in time following global reaction, considering your (bio)chemical & mathematical knowledge.,→ - Ensure all equations are (bio)chemically plausible and structurally consistent with known kin...

  74. [101]

    additional_info

    Additional information: {info["additional_info"]}. * EXAMPLE: The following example illustrates the expected output structure and level of detail. It is NOT intended to define the type of mechanism or domain-specific chemistry to be used in the actual problem. ,→ ,→ BEGINNING ...

  75. [102]

    Global reaction: CO + 1/2 O2→CO2

  76. [103]

    Carbon dioxide (CO2): CO2

    Products names: ("Carbon dioxide (CO2): CO2"). Reactants names: ("Carbon monoxide (CO): CO", "Oxygen (O2): O2"). Answer: A plausible new kinetic equation is r = (k6 * KCO * CO * (KO2 * O2)^(1/2)) / (1 + KCO * CO + KO2 * O2),→ This expression differs from all previous candidate...

  77. [104]

    Products and reactants names (if known)

  78. [105]

    The current 5 best candidate rate laws found by Symbolic Regression, with their AIC value and an explanation of why they could be plausible for the global reaction. ,→ ,→ * TASK: - Propose three kinetic equations for the following global reaction, considering the available exp...

  79. [106]

    balanced_reaction

    Global reaction: {info["balanced_reaction"]}

  80. [107]

    products

    Products names: ({info["products"]}). Reactants names: ({info["reactants"]})

  81. [108]

    Experimental data: {data}

  82. [109]

    best_candidates_analysis

    Analysis of current best candidate rate laws: {info["best_candidates_analysis"]}.,→ * EXAMPLE: The following example illustrates the expected output structure and level of detail. It is NOT intended to define the type of mechanism or domain-specific chemistry to be used in the...

  83. [110]

    53 LLM-guided Symbolic Regression for Kinetic Model Discovery

    Global reaction: CO + 1/2 O2→CO2. 53 LLM-guided Symbolic Regression for Kinetic Model Discovery

  84. [111]

    Carbon dioxide (CO2): CO2

    Products names: ("Carbon dioxide (CO2): CO2"). Reactants names: ("Carbon monoxide (CO): CO", "Oxygen (O2): O2")

  85. [112]

    Analysis of current best candidate rate laws: - (k1 * KCO * CO * KO2 * O2) / (1 + KCO * CO + KO2 * O2)^2 (AIC = 95.2) -> Plausible Langmuir-Hinshelwood mechanism with competitive adsorption.,→ - (k2 * CO * O2) / (1 + KCO * CO) (AIC = 101.8)→CO adsorption controls surface cover...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.