Pith. sign in

REVIEW 3 major objections 6 minor 59 references

Data-Efficient Symbolic Regression via Foundation Model Distillation

T0 review · 3 major / 6 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 EQUATE turns discrete equation search into continuous embedding-space optimization, guided by a learned fitness evaluator, and reports better equations across three benchmarks at low complexity.

desk verdict Novel fine-tuning framework for symbolic-regression foundation models with promising gains on standard benchmarks, but a load-bearing ambiguity about whether the search stops on test labels makes the headline results unverifiable as reported. read the letter →

arxiv 2508.19487 v1 pith:6TPIMHHA submitted 2025-08-27 cs.LG cs.AI

Wangyang Ying , Jinghan Zhang , Haoyue Bai , Nanxu Gong , Xinyuan Wang , Kunpeng Liu , Chandan K. Reddy , Yanjie Fu This is my paper · ORCID
classification cs.LGcs.AI
keywords symbolicregressionequationdiscoveryfoundationmodeldistillationembeddingspaceoptimizationevaluator-guidedsearchtransferlearningdata-efficienttransformers
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 generic pre-trained equation-discovery transformer, which by itself often transfers poorly to a single small domain-specific dataset, can be adapted to that dataset through a distillation-style fine-tuning loop. The proposed framework, EQUATE, samples many small subsets of the target data, uses the backbone to generate candidate equations, labels each data–equation pair by a $R^2$-score plus a simplicity term, and learns a shared embedding space for numerical data and symbolic equations. At inference it replaces beam search or sampling with gradient ascent in that embedding space, guided by a learned evaluator, then decodes the optimized embedding into the final equation. On the Feynman, Strogatz, and black-box benchmarks the paper reports consistent gains over the backbone and most comparison methods, with the largest gains on Strogatz where the backbone is weakest. If correct, this is a practical route to data-efficient symbolic regression: small annotated datasets become enough to specialize a large generative model without retraining it from scratch.

What carries the argument

The load-bearing object is the fused embedding of a data–equation pair, $E_f = E_n + \text{context}$, built from the pre-trained transformer's data embedding $E_n$ and an LSTM equation embedding $E_s$ through scaled-dot-product attention. The evaluator is a small MLP that regresses the quantity $r = (1-\sum_i(\tilde{y}_i - y_i)^2/\sum_i(\tilde{y}_i-\bar{y})^2) + \lambda\exp(-l(\tilde{f})/L)$, combining $R^2$ fitness with a simplicity penalty; its gradient with respect to $E_f$ supplies the search update $E_f^+ = E_f + \eta\,\partial r/\partial E_f$. During training, the reconstruction loss of the partially frozen decoder and the evaluator's mean-square error are jointly minimized, so the embedding space is shaped to make fitness locally smooth and predictable. At inference, the optimized embedding is decoded autoregressively into a prefix-notation equation. The mechanism's point is to supply a task-specific gradient direction, replacing token-level likelihood, which rewards equations that look like the pretraining data, with a continuous signal that rewards equations that fit this dataset.

What would settle it

Log the evaluator's predicted fitness and the actual $R^2$ of the decoded equation at every gradient step on a fixed benchmark equation; systematic divergence between predicted and actual scores as the search moves away from its starting embedding would show the evaluator is miscalibrated. A complementary check is to replace the evaluator with the true $R^2$ computed on the same sampled subset and require EQUATE to match that oracle and beat a random walk in the same embedding space by roughly the claimed margin.

Watch

Extended reading notes

Core claim

EQUATE's central claim is that discrete equation search can be reformulated as continuous optimization in a shared symbolic-numeric embedding space, and that this reformulation cures the negative-transfer problem of foundation models in low-data symbolic regression. The paper argues that fine-tuning a partially frozen transformer encoder-decoder alongside an LSTM equation encoder and an evaluator produces a space in which fitness is locally predictable, so that gradient steps on the evaluator's prediction with respect to the fused embedding move toward regions whose decoded equations fit the target data. The reported results support the claim that this beats the backbone's own decoding strategies as well as most genetic-programming and deep baselines, improves resilience to label noise, and does not inflate equation complexity or inference cost. The paper also claims that the best fine-tuning recipe keeps most pretrained weights frozen, updating only the equation encoder and the final decoder layer, and that both the symbolic alignment and the evaluator-guided search are necessary for the full gain.

Load-bearing premise

The framework assumes that the learned evaluator predicts true data–equation fitness accurately in exactly the region of embedding space that gradient ascent visits; if its predictions are miscalibrated there, the search can be steered toward embeddings whose decoded equations are worse than the backbone's own outputs.

Editorial extensions

If this is right

  • On Feynman, EQUATE with sampling raises the proportion of equations reaching test $R^2 > 0.99$ from 0.815 to 0.874 and average $R^2$ from 0.9730 to 0.9833 relative to the same decoding strategy of the backbone.
  • On Strogatz, the same variant raises $R^2 > 0.99$ from 0.357 to 0.786 and average $R^2$ from 0.8156 to 0.9818, showing that the method can deliver its largest gain exactly where the backbone transfers poorly.
  • The frozen fine-tuning recipe outperforms full fine-tuning, LoRA, and random initialization on all three benchmarks, supporting the paper's claim that preserving pretrained priors is central to the gain.
  • Ablations show that removing the equation encoder hurts the high-accuracy $R^2 > 0.99$ metric more than average $R^2$, while removing embedding-space optimization keeps average $R^2$ largely intact but loses high-accuracy cases; both components are needed for the best results.

Reading between the lines

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

  • The authors leave implicit that the same alignment-and-evaluator loop is a general template: any pre-trained sequence model that emits structured discrete objects, such as feature transformations or program sketches, could be adapted to a new dataset whenever a cheap numeric fitness exists.
  • Because 90 percent of training candidates are generated by the backbone itself, the search is confined by the backbone's candidate distribution; increasing the random share or adding genetic-programming candidates would test whether the evaluator can guide the search beyond that distribution.
  • The paper does not compare evaluator-guided gradient ascent with simply retrieving the best-known observed embedding by fitness; that comparison would separate the contribution of continuous search from the contribution of the aligned embedding space itself.
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

3 major / 6 minor

Summary. The paper proposes EQUATE, a fine-tuning framework for the end-to-end symbolic regression transformer E2E. EQUATE constructs a task-specific fine-tuning set by sampling subsets of the given dataset and generating candidate equations with the pretrained backbone, then learns a shared symbolic-numeric embedding space with a data encoder, an LSTM equation encoder, an attention-based fusion module, and an MLP evaluator that predicts a fitness-and-simplicity score. At inference, it performs gradient ascent in the fused embedding space under evaluator guidance, decodes the optimized embedding into an equation, and stops when a candidate with R2 > 0.99 is found or 100 candidates are generated. Experiments on Feynman, Strogatz, and black-box benchmarks report improved R2 and R2 > 0.99 over the E2E backbone and several SRBench baselines, together with ablations, a noise-robustness study, and an analysis of the accuracy-simplicity trade-off.

Significance. If the reported results withstand scrutiny, the paper makes a useful practical contribution: it shows a lightweight way to adapt a pretrained equation-discovery foundation model to small domain-specific datasets, combining symbolic and numeric modalities in a shared embedding space and using a learned evaluator to steer a continuous search. The framing of equation discovery as embedding search rather than pure token decoding is well motivated, and the paper explicitly evaluates complexity, inference time, and noise robustness, which are all relevant to practical symbolic regression. The main strengths are the clear problem statement, the use of publicly available E2E checkpoints and SRBench benchmarks, and the component ablations that isolate the equation encoder and embedding-space optimization. However, two evaluation-protocol issues currently prevent full support of the central claim: the data split used for the stopping criterion is unspecified, and the printed R2 formula is non-standard. These issues must be resolved before the performance comparisons can be taken as evidence for the proposed mechanism.

major comments (3)
  1. [Section 4.1] The paper never states which data split is used for the R2 that triggers early stopping: 'The search stops once an equation with R2 > 0.99 is found or the maximum number of candidates reaches 100.' If this R2 is computed on the test split, EQUATE is selecting the best of up to 100 candidates using the exact labels used for evaluation, whereas the E2E baselines decode without this test feedback, so the large improvements in Table 1 (e.g., Strogatz R2 > 0.99 rising from 0.357 to 0.786) could be test-set selection artifacts rather than evidence for the embedding-search mechanism. The authors must specify the split explicitly and, if test data are used for stopping, rerun the comparison with stopping decided on training or validation data only, or give all baselines the same candidate-selection protocol.
  2. [Equation (3) and Section 4.1] The R2 formula is written with the equation-estimated labels in the denominator: r = 1 - Σ(\tilde{y}_i - y_i)^2 / Σ(\tilde{y}_i - \bar{y})^2, where \bar{y} is the average true label. The standard coefficient of determination uses Σ(y_i - \bar{y})^2 in the denominator. As printed, the quantity used for evaluator training, the early-stopping criterion, and the reported metric is not standard R2 and can differ substantially from the SRBench baselines' definition, especially for poorly fitted equations. Please replace the denominator with Σ(y_i - \bar{y})^2 (if that was intended) and confirm that all R2 values in Tables 1-2 and Figures 3-6 are computed with the standard definition; otherwise, the reported comparisons are not on the same scale as the baselines.
  3. [Section 3.4 and Section 4.5] The gradient-ascent search moves embeddings according to the evaluator's predicted fitness, but the paper reports no validation that the evaluator's predictions correlate with actual R2 in the embedding region visited by the search. Without such calibration, the improvement attributed to 'evaluator-guided embedding space optimization' could instead arise from the candidate-selection or early-stopping rule. Please report a calibration analysis (e.g., predicted versus true fitness on held-out data-equation pairs, and actual R2 along the search trajectory) and, if needed, re-evaluate the ablation in Section 4.5 with this evidence.
minor comments (6)
  1. [Appendix A.6] The phrase 'the search step size to a maximum of 20 steps' conflates the gradient step size with the number of gradient steps; please specify the numerical value of the step size η and the number of steps separately.
  2. [Section 4.4] The noise-robustness experiment does not state the noise distribution, whether noise is added only to training labels, how many repetitions were averaged, or whether error bars are shown; please provide these details.
  3. [Table 2] The Black-box column for the frozen row lists a dash for R2 > 0.99; please explain why this metric is omitted for that condition.
  4. [Figure 3] Figure 3(b) labels the baseline as 'SINP', while the text and baseline list use 'SNIP'; please correct the label.
  5. [Section 4.5] The ablation figure reports only point estimates without numerical values or error bars; please report the exact metrics and their variance across runs.
  6. [Section 4.2] The claim that EQUATE maintains complexity 'either lower than or comparable' to the backbone is not supported for Feynman, where EQUATE-Sampling complexity is 57.76 versus 53.61 for E2E-Sampling; please qualify this statement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: EQUATE's search and evaluation are not equivalent to its training inputs by construction.

full rationale

The paper's derivation chain is self-contained rather than circular. The fine-tuning training set is prepared by sampling subsets and generating candidate equations with the pretrained backbone, but the fitness labels used for the evaluator are computed from Eq. (3) as a discrepancy between gold labels and equation-estimated labels, so the supervision is grounded in the observed data rather than in the model's own predictions. The evaluator is a learned surrogate; the gradient-ascent search in the shared embedding space can move to embeddings that are not simply the training instances, and the final equations are decoded and scored against held-out test datapoints (Section 4.1). No fitted parameter is renamed as a prediction, and no equation in the paper is shown to equal a training-set equation by construction. The backbone E2E [23] is an external pretrained model; the authors' self-citations in Related Work are application or baseline references and are not load-bearing for the central claim. No uniqueness theorem or ansatz is imported from the authors' prior work. The only notable ambiguity is that Section 4.1 does not specify which data split is used in the early-stopping rule ('The search stops once an equation with R2 > 0.99 is found'); if that R2 were computed on the test split, the comparison would be a test-selection artifact, but that is a benchmarking-correctness concern rather than a circular derivation, and the text does not state that the stopping R2 is a test R2. The stated Limitation (input dimensionality capped at 10) is a scope restriction and does not affect circularity.

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

The central claim rests on the transferability of the E2E foundation model, the reliability of the learned evaluator as a fitness surrogate, and a set of hand-chosen hyperparameters (alpha, beta, lambda, search steps). No new physical entities are introduced. The most fragile component is the evaluator, since the entire search step depends on its calibration.

free parameters (3)
  • Loss weights alpha and beta = alpha = 0.05, beta = 100
    Chosen by hand to balance evaluator MSE and reconstruction NLL; not derived from data.
  • Complexity trade-off parameter lambda = Not reported for main results; varied in {0, 0.1, 0.5, 1} in Appendix C
    Controls the simplicity term in the evaluator objective; the default value is not stated, so the main results depend on an unreported choice.
  • Gradient search step size eta and number of steps = eta not reported; steps up to 20
    The step size is not specified, so the search behavior is not fully reproducible; the step count is set to 20.
assumptions (3)
  • domain assumption The pre-trained E2E encoder and decoder weights provide useful priors for the target datasets.
    Used throughout; Section 3.6 freezes the data encoder and most of the decoder to transfer knowledge; Table 2 shows random initialization yields negative R2.
  • domain assumption The evaluator predicts true data-equation fitness reliably over the embedding region explored by gradient ascent.
    Section 3.4 uses the evaluator's gradient to move embeddings; no validation of evaluator accuracy against actual R2 is reported.
  • ad hoc to paper The stopping R2 in Section 4.1 is computed on training data, not held-out test data.
    The paper does not state which split is used for the stopping criterion; if test data is used, the evaluation is circular.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Data-Efficient Symbolic Regression via Foundation Model Distillation." pith.science (2026). https://pith.science/paper/6TPIMHHA

@misc{pith2026250819487,
  author       = {Pith},
  title        = {Pith review of: Data-Efficient Symbolic Regression via Foundation Model Distillation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6TPIMHHA}},
  note         = {Machine review of arXiv:2508.19487}
}
read the original abstract

Discovering interpretable mathematical equations from observed data (a.k.a. equation discovery or symbolic regression) is a cornerstone of scientific discovery, enabling transparent modeling of physical, biological, and economic systems. While foundation models pre-trained on large-scale equation datasets offer a promising starting point, they often suffer from negative transfer and poor generalization when applied to small, domain-specific datasets. In this paper, we introduce EQUATE (Equation Generation via QUality-Aligned Transfer Embeddings), a data-efficient fine-tuning framework that adapts foundation models for symbolic equation discovery in low-data regimes via distillation. EQUATE combines symbolic-numeric alignment with evaluator-guided embedding optimization, enabling a principled embedding-search-generation paradigm. Our approach reformulates discrete equation search as a continuous optimization task in a shared embedding space, guided by data-equation fitness and simplicity. Experiments across three standard public benchmarks (Feynman, Strogatz, and black-box datasets) demonstrate that EQUATE consistently outperforms state-of-the-art baselines in both accuracy and robustness, while preserving low complexity and fast inference. These results highlight EQUATE as a practical and generalizable solution for data-efficient symbolic regression in foundation model distillation settings.

Figures

Figures reproduced from arXiv: 2508.19487 by the authors.

Figure 1
Figure 1. FKT transfers generic knowledge to specific do [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the EQUATE framework: a fine-tuning architecture that embeds data and symbolic equations, aligns [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Performance comparison of EQUATE and all SRBench algorithms for Feynman and Strogatz datasets. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Performance in terms of robustness to noise. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Verify the key components of the equation encoder [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 7
Figure 7. Figure 7: Complexity check on Strogatz dataset. EQUATE [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 24 canonical work pages

  1. [1]

    Dimitrios Angelis, Filippos Sofos, and Theodoros E Karakasidis. 2023. Artificial intelligence in physical sciences: Symbolic regression trends and perspectives. Archives of Computational Methods in Engineering 30, 6 (2023), 3845–3865

    work page 2023

  2. [2]

    Ignacio Arnaldo, Krzysztof Krawiec, and Una-May O’Reilly. 2014. Multiple regression genetic programming. In Proceedings of the 2014 annual conference on genetic and evolutionary computation . 879–886

    work page 2014

  3. [3]

    Douglas Adriano Augusto and Helio JC Barbosa. 2000. Symbolic regression via genetic programming. In Proceedings. Vol. 1. Sixth Brazilian symposium on neural networks. IEEE, 173–178

    work page 2000

  4. [4]

    Haoyue Bai, Guodong Chen, Wangyang Ying, Xinyuan Wang, Nanxu Gong, Sixun Dong, Giulia Pedrielli, Haoyu Wang, Haifeng Chen, and Yanjie Fu. 2025. Brownian Bridge Augmented Surrogate Simulation and Injection Planning for Geological CO _2 Storage. arXiv preprint arXiv:2505.18204 (2025)

    arXiv 2025

  5. [5]

    Luca Biggio, Tommaso Bendinelli, Alexander Neitz, Aurelien Lucchi, and Giambat- tista Parascandolo. 2021. Neural symbolic regression that scales. In International Conference on Machine Learning . Pmlr, 936–945

    work page 2021

  6. [6]

    Bogdan Burlacu, Gabriel Kronberger, and Michael Kommenda. 2020. Operon C++: an efficient genetic programming framework for symbolic regression. In Proceedings of the 2020 Genetic and Evolutionary Computation Conference Com- panion (Cancún, Mexico) (GECCO ’20). Association for Computing Machinery, New York, NY, USA, 1562–1570. https://doi.org/10.1145/33779...

    arXiv 2020

  7. [7]

    Birkan Can and Cathal Heavey. 2011. Comparison of experimental designs for simulation-based symbolic regression of manufacturing systems. Computers & Industrial Engineering 61, 3 (2011), 447–462

    work page 2011

  8. [8]

    William La Cava, Patryk Orzechowski, Bogdan Burlacu, Fabrício Olivetti de França, Marco Virgolin, Ying Jin, Michael Kommenda, and Jason H. Moore. 2021. Contemporary Symbolic Regression Methods and their Relative Performance. arXiv:2107.14351 [cs.NE] https://arxiv.org/abs/2107.14351

    arXiv 2021

Show all 59 references
  1. [9]

    Oscar Claveria, Enric Monte, and Salvador Torra. 2017. Assessment of the effect of the financial crisis on agents’ expectations through symbolic regression. Applied Economics Letters 24, 9 (2017), 648–652

    2017

  2. [10]

    Miles Cranmer, Alvaro Sanchez Gonzalez, Peter Battaglia, Rui Xu, Kyle Cranmer, David Spergel, and Shirley Ho. 2020. Discovering symbolic models from deep learning with inductive biases. Advances in neural information processing systems 33 (2020), 17429–17442

    2020

  3. [11]

    Fabricio Olivetti de Franca and Guilherme Seidyo Imai Aldeia. 2021. Interaction– transformation evolutionary algorithm for symbolic regression. Evolutionary computation 29, 3 (2021), 367–390

    2021

  4. [12]

    Stéphane d’Ascoli, Pierre-Alexandre Kamienny, Guillaume Lample, and Francois Charton. 2022. Deep symbolic regression for recurrence prediction. In Interna- tional Conference on Machine Learning . PMLR, 4520–4536

    2022

  5. [13]

    Angela Fan, Mike Lewis, and Yann Dauphin. 2018. Hierarchical neural story generation. arXiv preprint arXiv:1805.04833 (2018)

    2018 arXiv

  6. [14]

    Nanxu Gong, Wangyang Ying, Dongjie Wang, and Yanjie Fu. 2025. Neuro- symbolic embedding for short and effective feature selection via autoregressive generation. ACM Transactions on Intelligent Systems and Technology 16, 2 (2025), 1–21

    2025

  7. [15]

    Alex Graves. 2012. Sequence transduction with recurrent neural networks. arXiv preprint arXiv:1211.3711 (2012)

    2012 arXiv

  8. [16]

    Gustafson, E.K

    S. Gustafson, E.K. Burke, and N. Krasnogor. 2005. On improving genetic program- ming for symbolic regression. In2005 IEEE Congress on Evolutionary Computation, Vol. 1. 912–919 Vol.1. https://doi.org/10.1109/CEC.2005.1554780

    2005

  9. [17]

    Christian Haider, Fabricio Olivetti de Franca, Bogdan Burlacu, and Gabriel Kron- berger. 2023. Shape-constrained multi-objective genetic programming for sym- bolic regression. Applied Soft Computing 132 (2023), 109855

    2023

  10. [18]

    Samuel Holt, Zhaozhi Qian, and Mihaela van der Schaar. 2023. Deep generative symbolic regression. arXiv preprint arXiv:2401.00282 (2023)

    2023 arXiv

  11. [19]

    Xuanming Hu, Dongjie Wang, Wangyang Ying, and Yanjie Fu. 2024. Reinforce- ment Feature Transformation for Polymer Property Performance Prediction. In Proceedings of the 33rd ACM International Conference on Information and Knowl- edge Management. 4538–4545

    2024

  12. [20]

    Menghao Huo, Kuan Lu, Yuxiao Li, Qiang Zhu, and Zhenrui Chen. 2025. Ct- patchtst: Channel-time patch time-series transformer for long-term renewable energy forecasting. arXiv preprint arXiv:2501.08620 (2025)

    2025

  13. [21]

    Menghao Huo, Kuan Lu, Qiang Zhu, and Zhenrui Chen. 2025. Enhancing cus- tomer contact efficiency with graph neural networks in credit card fraud detection workflow. arXiv preprint arXiv:2504.02275 (2025)

    2025

  14. [22]

    Ying Jin, Weilin Fu, Jian Kang, Jiadong Guo, and Jian Guo. 2019. Bayesian symbolic regression. arXiv preprint arXiv:1910.08892 (2019)

    2019 arXiv

  15. [23]

    Pierre-Alexandre Kamienny, Stéphane d’Ascoli, Guillaume Lample, and François Charton. 2022. End-to-end symbolic regression with transformers. Advances in Neural Information Processing Systems 35 (2022), 10269–10281

    2022

  16. [24]

    Samuel Kim, Peter Y Lu, Srijon Mukherjee, Michael Gilbert, Li Jing, Vladimir Čeperić, and Marin Soljačić. 2020. Integration of neural network-based symbolic regression in deep learning for scientific discovery. IEEE transactions on neural networks and learning systems 32, 9 (2...

    2020

  17. [25]

    William La Cava, Kourosh Danai, and Lee Spector. 2016. Inference of compact nonlinear dynamic models by epigenetic local search. Engineering Applications of Artificial Intelligence 55 (2016), 292–306

    2016

  18. [26]

    William La Cava, Tilak Raj Singh, James Taggart, Srinivas Suri, and Jason H Moore

  19. [27]

    William La Cava, Lee Spector, and Kourosh Danai. 2016. Epsilon-lexicase selec- tion for regression. In Proceedings of the Genetic and Evolutionary Computation Conference 2016. 741–748

    2016

  20. [28]

    William G La Cava, Paul C Lee, Imran Ajmal, Xiruo Ding, Priyanka Solanki, Jordana B Cohen, Jason H Moore, and Daniel S Herman. 2023. A flexible symbolic regression method for constructing interpretable clinical prediction models. NPJ Digital Medicine 6, 1 (2023), 107

    2023

  21. [29]

    Trent McConaghy. 2011. FFX: Fast, scalable, deterministic symbolic regression technology. In Genetic Programming Theory and Practice IX . Springer, 235–260

    2011

  22. [30]

    Kazem Meidani, Parshin Shojaee, Chandan K Reddy, and Amir Barati Farimani

  23. [31]

    Terrell Mundhenk, Mikel Landajuela, Ruben Glatt, Claudio P Santiago, Brenden K Petersen, et al. 2021. Symbolic regression via deep reinforcement learning en- hanced genetic programming seeding. Advances in Neural Information Processing Systems 34 (2021), 24912–24923

    2021

  24. [32]

    T Nathan Mundhenk, Mikel Landajuela, Ruben Glatt, Claudio P Santiago, Daniel M Faissol, and Brenden K Petersen. 2021. Symbolic regression via neural- guided genetic programming population seeding. arXiv preprint arXiv:2111.00053 (2021)

    2021 arXiv

  25. [33]

    Olson, William La Cava, Patryk Orzechowski, Ryan J

    Randal S. Olson, William La Cava, Patryk Orzechowski, Ryan J. Urbanowicz, and Jason H. Moore. 2017. PMLB: A Large Benchmark Suite for Machine Learning Evaluation and Comparison. arXiv:1703.00512 [cs.LG] https://arxiv.org/abs/ 1703.00512

    2017 arXiv

  26. [34]

    Brenden K Petersen, Mikel Landajuela, T Nathan Mundhenk, Claudio P Santiago, Soo K Kim, and Joanne T Kim. 2019. Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients. arXiv preprint arXiv:1912.04871 (2019)

    2019 arXiv

  27. [35]

    Michael D Schmidt and Hod Lipson. 2010. Age-fitness pareto optimization. In Proceedings of the 12th annual conference on Genetic and evolutionary computation . 9 543–544

    2010

  28. [36]

    Parshin Shojaee, Kazem Meidani, Amir Barati Farimani, and Chandan Reddy

  29. [37]

    Parshin Shojaee, Kazem Meidani, Shashank Gupta, Amir Barati Farimani, and Chandan K Reddy. 2024. Llm-sr: Scientific equation discovery via programming with large language models. arXiv preprint arXiv:2404.18400 (2024)

    2024 arXiv

  30. [38]

    Fangzheng Sun, Yang Liu, Jian-Xun Wang, and Hao Sun. 2022. Symbolic physics learner: Discovering governing equations via monte carlo tree search. arXiv preprint arXiv:2205.13134 (2022)

    2022 arXiv

  31. [39]

    Advances in Neural Information Processing Systems 36 (2023), 45907–45919

    Transformer-based planning for symbolic regression. Advances in Neural Information Processing Systems 36 (2023), 45907–45919

    2023

  32. [40]

    Silviu-Marian Udrescu and Max Tegmark. 2020. AI Feynman: A physics-inspired method for symbolic regression. Science Advances 6, 16 (2020), eaay2631

    2020

  33. [41]

    Nguyen Quang Uy, Nguyen Xuan Hoai, Michael O’Neill, Robert I McKay, and Edgar Galván-López. 2011. Semantically-based crossover in genetic program- ming: application to real-valued symbolic regression. Genetic Programming and Evolvable Machines 12 (2011), 91–119

    2011

  34. [42]

    Silviu-Marian Udrescu, Andrew Tan, Jiahai Feng, Orisvaldo Neto, Tailin Wu, and Max Tegmark. 2020. AI Feynman 2.0: Pareto-optimal symbolic regression exploiting graph modularity. arXiv:2006.10782 [cs.LG] https://arxiv.org/abs/ 2006.10782

    2020 arXiv

  35. [43]

    Marco Virgolin, Tanja Alderliesten, and Peter AN Bosman. 2019. Linear scal- ing with and within semantic backpropagation-based genetic programming for symbolic regression. In Proceedings of the genetic and evolutionary computation conference. 1084–1092

    2019

  36. [44]

    Marco Virgolin, Tanja Alderliesten, Cees Witteveen, and Peter AN Bosman

  37. [45]

    Mojtaba Valipour, Bowen You, Maysum Panju, and Ali Ghodsi. 2021. Symbol- icgpt: A generative transformer model for symbolic regression. arXiv preprint arXiv:2106.14131 (2021)

    2021 arXiv

  38. [46]

    Sam Wiseman and Alexander M Rush. 2016. Sequence-to-sequence learning as beam-search optimization. arXiv preprint arXiv:1606.02960 (2016)

    2016 arXiv

  39. [47]

    Wangyang Ying, Haoyue Bai, Nanxu Gong, Xinyuan Wang, Sixun Dong, Haifeng Chen, and Yanjie Fu. 2025. Bridging the domain gap in equation distillation with reinforcement feedback. arXiv preprint arXiv:2505.15572 (2025)

    2025 arXiv

  40. [48]

    Wangyang Ying, Haoyue Bai, Kunpeng Liu, and Yanjie Fu. 2024. Topology-aware Reinforcement Feature Space Reconstruction for Graph Data. arXiv preprint arXiv:2411.05742 (2024)

    2024 arXiv

  41. [49]

    Yiqun Wang, Nicholas Wagner, and James M Rondinelli. 2019. Symbolic regres- sion in materials science. MRS Communications 9, 3 (2019), 793–805

    2019

  42. [50]

    Wangyang Ying, Dongjie Wang, Xuanming Hu, Ji Qiu, Jin Park, and Yanjie Fu. 2024. Revolutionizing Biomarker Discovery: Leveraging Generative AI for Bio-Knowledge-Embedded Continuous Space Exploration. In Proceedings of the 33rd ACM International Conference on Information and Kn...

    2024

  43. [51]

    Wangyang Ying, Dongjie Wang, Xuanming Hu, Yuanchun Zhou, Charu C Ag- garwal, and Yanjie Fu. 2024. Unsupervised generative feature transformation via graph contrastive pre-training and multi-objective fine-tuning. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Di...

    2024

  44. [52]

    Wangyang Ying, Dongjie Wang, Kunpeng Liu, Leilei Sun, and Yanjie Fu. 2023. Self-optimizing feature generation via categorical hashing representation and hierarchical reinforcement crossing. In 2023 IEEE International Conference on Data Mining (ICDM). IEEE, 748–757

    2023

  45. [53]

    Wangyang Ying, Dongjie Wang, Haifeng Chen, and Yanjie Fu. 2024. Feature selection as deep sequential generative learning. ACM Transactions on Knowledge Discovery from Data 18, 9 (2024), 1–21

    2024

  46. [54]

    Jinghui Zhong, Liang Feng, Wentong Cai, and Yew-Soon Ong. 2018. Multifactorial genetic programming for symbolic regression problems. IEEE transactions on systems, man, and cybernetics: systems 50, 11 (2018), 4492–4505

    2018

  47. [55]

    Qiang Zhu, Kuan Lu, Menghao Huo, and Yuxiao Li. 2025. Image-to-image trans- lation with diffusion transformers and clip-based image conditioning. arXiv preprint arXiv:2505.16001 (2025). A Experimental Settings Implementation Details. We leverage the state-of-the-art open- sour...

    2025

  48. [57]

    Michael Zhang, Samuel Kim, Peter Y Lu, and Marin Soljačić. 2023. Deep learning and symbolic regression for discovering parametric equations. IEEE Transactions on Neural Networks and Learning Systems (2023)

    2023

  49. [2017]

    In Proceedings of the Genetic and Evolutionary Computation Conference

    Scalable genetic programming by gene-pool optimal mixing and input- space entropy-based building-block learning. In Proceedings of the Genetic and Evolutionary Computation Conference. 1041–1048

  50. [2018]

    arXiv preprint arXiv:1807.00981 (2018)

    Learning concise representations for regression by evolving networks of trees. arXiv preprint arXiv:1807.00981 (2018)

    2018 arXiv

  51. [2023]

    arXiv preprint arXiv:2310.02227 (2023)

    Snip: Bridging mathematical symbolic and numeric realms with unified pre-training. arXiv preprint arXiv:2310.02227 (2023)

    2023 arXiv

Pith tools

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