arxiv: 2604.06998 · v1 · submitted 2026-04-08 · ❄️ cond-mat.other

Recognition: unknown

Identifying Topological Invariants of Non-Hermitian Systems via Domain-Adaptive Multimodal Model for Mathematics

Jiuchun Meng , Lichao Sun , Xiumei Wang , Dandan Zhu , Xingping Zhou

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:11 UTC · model grok-4.3

classification ❄️ cond-mat.other

keywords non-Hermitian systemstopological invariantsskin effectmultimodal modellarge language modelseigenvalues and eigenvectorsmomentum space

0 comments

The pith

A domain-adaptive multimodal model with Qwen Math backbone identifies topological invariants in non-Hermitian systems from eigenvalues and eigenvectors.

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

The paper proposes a framework that feeds eigenvalues and eigenvectors of the Hamiltonian in momentum space into a domain-adaptive multimodal model for mathematics to determine topological invariants of non-Hermitian systems. This model uses Qwen Math as its backbone to improve mathematical reasoning and precision. The approach targets the challenge of implementing generalized Brillouin zone or amoeba formulations in high dimensions, where conventional algorithms struggle. By treating the two inputs as separate modalities, the model aims to extract the invariants without heavy reliance on labeled data or extra physics rules. If the method works, it offers a reusable template for applying large language models to similar identification tasks in condensed-matter physics.

Core claim

The central claim is that eigenvalues and eigenvectors alone, processed as dual modalities through a multimodal model built on the Qwen Math backbone, suffice to identify topological invariants in non-Hermitian systems and thereby supply a practical paradigm for future LLM-based studies of these invariants.

What carries the argument

The domain-adaptive multimodal model for mathematics that takes eigenvalues and eigenvectors of the momentum-space Hamiltonian as separate input modalities and uses Qwen Math integration to perform the identification.

If this is right

High-dimensional non-Hermitian systems become computationally tractable for invariant identification where generalized Brillouin zone methods are difficult to implement.
The need for extensive labeled datasets or hand-crafted domain constraints is reduced.
Large language models gain a demonstrated route for extracting topological information directly from spectral data.
The same multimodal architecture can serve as a template for other invariant-identification problems in physics.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The method could be tested on standard non-Hermitian models such as the two-dimensional Hatano-Nelson lattice to produce quick validation metrics.
Combining the model output with symbolic verification steps might tighten the precision of the extracted invariants.
If the approach succeeds on non-Hermitian skin-effect cases, analogous multimodal inputs could be explored for Hermitian topological phases.

Load-bearing premise

Eigenvalues and eigenvectors alone, without domain-specific physics constraints or large amounts of labeled training data, contain everything the model needs to correctly identify the invariants.

What would settle it

Apply the trained model to a concrete high-dimensional non-Hermitian Hamiltonian whose topological invariant is already known from analytic calculation and check whether the model's output matches that known value.

read the original abstract

The emergence of the non-Hermitian skin effect, distinguished by the exponential localization of bulk states onto boundaries in open systems, has redefined the conventional band theory. It can be established through the generalized Brillouin zone framework, the amoeba formulation or generalized Fermi surface in the different dimensions. However, its algorithmic implementation is still challenging in the high-dimensional cases. The large language models (LLM), functioning as the new paradigm in machine learning, can help to tack scientific problems. Here, we propose a framework composed by domain-adaptive Multimodal model for mathematics to identify topological invariants. We feed the eigenvalues and eigenvectors of the Hamiltonian in momentum space into our model as two input modalities. The Qwen Math is integrated as the backbone of the multimodal model, significantly enhancing its mathematical understanding capability and computational precision. Our results provide a paradigm for future studies on topological invariants identification via LLMs.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note private letter to a colleague

The paper proposes feeding eigenvalues and eigenvectors into a Qwen-Math multimodal model to identify non-Hermitian topological invariants but supplies no training details, benchmarks, or examples to show it works.

read the letter

The paper's main suggestion is to treat eigenvalues and eigenvectors as two input modalities for a domain-adapted multimodal LLM built on Qwen Math, with the goal of extracting topological invariants in non-Hermitian systems where high-dimensional cases make conventional algorithms slow. It frames this as a way around the challenges of generalized Brillouin zones, amoeba formulations, and skin-effect localization. The specific combination of multimodal inputs with a math-specialized LLM for this condensed-matter setting is the clearest new element. The motivation is stated plainly and the problem is real: standard methods do get expensive in higher dimensions. That part is useful to note. The rest of the manuscript, however, stays at the level of describing the architecture without showing any implementation. There is no account of how training data would be generated from Hamiltonians with known invariants, no loss function or training objective, no accuracy numbers on test cases, and no comparison against existing numerical methods. The claim that the model thereby provides a new paradigm therefore rests on an untested assertion that the two modalities plus the LLM's pretraining are informationally complete. The assumption that no additional non-Hermitian constraints need to be injected also goes unexamined. Readers working at the edge of machine learning and topological physics might find the outline worth a quick look as a prompt for their own experiments. Anyone expecting a working method or reproducible result will not get it here. The thinking on the motivation is clear, but the absence of evidence makes the central claim impossible to evaluate. I would not bring this to a reading group or cite it. It does not yet deserve peer review; the authors would need to add concrete training, validation on known invariants, and error analysis before it becomes a serious submission.

Referee Report

1 major / 2 minor

Summary. The manuscript proposes a domain-adaptive multimodal model for mathematics, with Qwen-Math as backbone, to identify topological invariants of non-Hermitian systems. Eigenvalues and eigenvectors of the Hamiltonian in momentum space are supplied as two input modalities; the approach is asserted to overcome algorithmic challenges in high-dimensional cases (e.g., generalized Brillouin zone or amoeba formulations of the skin effect) and to furnish a new paradigm for LLM-based invariant identification.

Significance. If the model were shown to extract invariants such as winding or Chern numbers reliably from eigendecomposition data alone, the work could supply a data-driven route to problems where conventional high-dimensional implementations remain difficult. The integration of a mathematics-specialized LLM backbone is a timely direction. At present, however, the absence of any training protocol, dataset, loss function, accuracy figures, or validation on known cases prevents any positive assessment of significance.

major comments (1)

[Abstract] Abstract: the assertion 'Our results provide a paradigm for future studies on topological invariants identification via LLMs' is unsupported; the manuscript contains no quantitative results, accuracy metrics, baseline comparisons, error analysis, training details, or validation examples on systems with known invariants (e.g., winding numbers under point-gap topology).

minor comments (2)

[Abstract] Abstract: 'can help to tack scientific problems' contains a typographical error ('tack' for 'tackle').
[Abstract] Abstract: 'composed by domain-adaptive Multimodal model' is grammatically imprecise ('composed of' is the conventional phrasing).

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive and detailed report. We agree that the current manuscript is a conceptual proposal for a domain-adaptive multimodal framework and does not yet contain the quantitative validation, training details, or benchmark results needed to support the abstract claim. We will perform a major revision to address this.

read point-by-point responses

Referee: [Abstract] Abstract: the assertion 'Our results provide a paradigm for future studies on topological invariants identification via LLMs' is unsupported; the manuscript contains no quantitative results, accuracy metrics, baseline comparisons, error analysis, training details, or validation examples on systems with known invariants (e.g., winding numbers under point-gap topology).

Authors: We fully agree that the assertion is currently unsupported. The manuscript outlines the architecture (eigenvalue and eigenvector modalities fed to a Qwen-Math backbone) and motivates its use for high-dimensional non-Hermitian problems where generalized Brillouin zone or amoeba methods become intractable, but it does not include implementation, training, or numerical tests. In the revised version we will add: (i) a dataset of known 1D and 2D non-Hermitian Hamiltonians with analytically computable invariants (winding numbers for point-gap topology and Chern numbers for line-gap cases), (ii) the precise training protocol, loss function, and multimodal fusion strategy, (iii) accuracy, precision, and error metrics on held-out test systems, and (iv) comparisons against conventional numerical diagonalization and existing machine-learning baselines. The abstract claim will be either substantiated by these results or appropriately qualified. revision: yes

Circularity Check

0 steps flagged

No circularity; proposed ML pipeline is self-contained as an unverified methodological suggestion

full rationale

The manuscript describes a framework that feeds eigenvalues and eigenvectors of non-Hermitian Hamiltonians into a Qwen-Math-backed multimodal model to identify topological invariants. No derivation, equation, or proof is supplied that reduces by construction to its own inputs, fitted parameters, or self-citations. The central claim is an assertion that the two modalities plus the backbone suffice, without any self-definitional loop, renamed empirical pattern, or load-bearing self-citation chain. Absence of training details, loss functions, or benchmarks is a gap in evidence, not a circular reduction. The approach remains an independent (if untested) proposal rather than a tautology.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on the assumption that the chosen input modalities suffice for the task and that the LLM backbone can generalize to identify invariants. No explicit free parameters, invented entities, or additional axioms are stated.

axioms (1)

domain assumption Eigenvalues and eigenvectors of the Hamiltonian in momentum space contain sufficient information to determine topological invariants.
This is implicit in the choice to feed these quantities as the two input modalities.

pith-pipeline@v0.9.0 · 5466 in / 1210 out tokens · 67588 ms · 2026-05-10T18:11:23.547652+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

73 extracted references · 15 canonical work pages · 10 internal anchors

[1]

Okuma and M

N. Okuma and M. Sato, Non-Hermitian topological phenomena: A review, Annu. Rev. Condens. Matter Phys. 14, 83 (2023)

2023
[2]

E. J. Bergholtz, J. C. Budich, and F. K. Kunst, Exceptional topology of non-Hermitian systems, Rev. Mod. Phys. 93, 015005 (2021)

2021
[3]

Zhou and D.-J

L. Zhou and D.-J. Zhang, Non-Hermitian Floquet topological matter—a review, Entropy 25, 1401 (2023)

2023
[4]

Yao and Z

S. Yao and Z. Wang, Edge states and topological invariants of non-Hermitian systems, Phys. Rev. Lett. 121, 086803 (2018)

2018
[5]

S. Yao, F. Song, and Z. Wang, Non-Hermitian Chern bands, Phys. Rev. Lett. 121, 136802 (2018)

2018
[6]

F. K. Kunst, E. Edvardsson, J. C. Budich, and E. J. Bergholtz, Biorthogonal bulk-boundary correspondence in non-Hermitian systems, Phys. Rev. Lett. 121, 026808 (2018)

2018
[7]

V. M. Martinez Alvarez, J. E. Barrios Vargas, and L. E. F. Foa Torres, Non-Hermitian robust edge states in one dimension: Anomalous localization and eigenspace condensation at exceptional points, Phys. Rev. B 97, 121401 (2018)

2018
[8]

C. H. Lee and R. Thomale, Anatomy of skin modes and topology in non-Hermitian systems, Phys. Rev. B 99, 201103 (2019)

2019
[9]

Yokomizo and S

K. Yokomizo and S. Murakami, Non-Bloch band theory of non-Hermitian systems, Phys. Rev. Lett. 123, 066404 (2019)

2019
[10]

Kawabata, K

K. Kawabata, K. Shiozaki, M. Ueda, and M. Sato, Symmetry and topology in non-Hermitian physics, Phys. Rev. X 9, 041015 (2019)

2019
[11]

Okuma, M

N. Okuma, M. Sato, and K. Shiozaki, Topological origin of non-Hermitian skin effects, Phys. Rev. Lett. 124, 086801 (2020)

2020
[12]

D. S. Borgnia, A. J. Kruchkov, and R.-J. Slager, Non-Hermitian boundary modes and topology, Phys. Rev. Lett. 124, 056802 (2020)

2020
[13]

L. Li, C. H. Lee, and J. Gong, Topological switch for non-Hermitian skin effect in cold-atom systems with loss, Phys. Rev. Lett. 124, 250402 (2020)

2020
[14]

Helbig, T

T. Helbig, T. Hofmann, S. Imhof, M. Pietryga, V. Michishita, S. Ozawa, T. Kiessling, M. Greiter, R. Thomale, and C. H. Lee, Generalized bulk-boundary correspondence in non-Hermitian topolectrical circuits, Nat. Phys. 16, 747 (2020)

2020
[15]

L. Xiao, T. Deng, K. Wang, G. Zhu, Z. Wang, W. Yi, and P. Xue, Non-Hermitian bulk-boundary correspondence in quantum dynamics, Nat. Phys. 16, 761 (2020)

2020
[16]

Weidemann, M

S. Weidemann, M. Kremer, T. Helbig, T. Hofmann, A. Stegmaier, M. Greiter, R. Thomale, and A. Szameit, Topological funneling of light, Science 368, 311 (2020)

2020
[17]

Zhang, Z

K. Zhang, Z. Yang, and C. Fang, Universal non-Hermitian skin effect in two and higher dimensions, Nat. Commun. 13, 2496 (2022)

2022
[18]

H.-Y. Wang, F. Song, and Z. Wang, Amoeba formulation of non-Bloch band theory in arbitrary dimensions, Phys. Rev. X 14, 021011 (2024)

2024
[19]

Zhang, C

K. Zhang, C. Shu, and K. Sun, Algebraic non-Hermitian skin effect and generalized Fermi surface formula in arbitrary dimensions, Phys. Rev. X 15, 031039 (2025)

2025
[20]

Zhang, H

P. Zhang, H. Shen, and H. Zhai, Machine learning topological invariants with neural networks, Phys. Rev. Lett. 120, 066401 (2018)

2018
[21]

Zhang, L.-Z

L.-F. Zhang, L.-Z. Tang, Z.-H. Huang, G.-Q. Zhang, W. Huang, and D.-W. Zhang, Machine learning topological invariants of non-Hermitian systems, Phys. Rev. A 103, 012419 (2021)

2021
[22]

Zhang, R

Y. Zhang, R. G. Melko, and E.-A. Kim, Machine learning quantum spin liquids with quasiparticle statistics, Phys. Rev. B 96, 245119 (2017)

2017
[23]

Yoshioka, Y

N. Yoshioka, Y. Akagi, and H. Katsura, Learning disordered topological phases by statistical recovery of symmetry, Phys. Rev. B 97, 205110 (2018)

2018
[24]

X.-Y. Dong, F. Pollmann, and X.-F. Zhang, Machine learning of quantum phase transitions, Phys. Rev. B 99, 121104 (2019)

2019
[25]

N. L. Holanda and M. A. R. Griffith, Machine learning topological phases in real space, Phys. Rev. B 102, 054107 (2020)

2020
[26]

Narayan and A

B. Narayan and A. Narayan, Machine learning non-Hermitian topological phases, Phys. Rev. B 103, 035413 (2021)

2021
[27]

Lian et al., Machine learning topological phases with a solid-state quantum simulator, Phys

W. Lian et al., Machine learning topological phases with a solid-state quantum simulator, Phys. Rev. Lett. 122, 210503 (2019)

2019
[28]

Mertz and R

T. Mertz and R. Valentí, Engineering topological phases guided by statistical and machine learning methods, Phys. Rev. Res. 3, 013132 (2021)

2021
[29]

Zhang, S

H. Zhang, S. Jiang, X. Wang, W. Zhang, X. Huang, X. Ouyang, Y. Yu, Y. Liu, D.-L. Deng, and L.-M. Duan, Experimental demonstration of adversarial examples in learning topological phases, Nat. Commun. 13, 4993 (2022)

2022
[30]

Arnold and F

J. Arnold and F. Schäfer, Replacing neural networks by optimal analytical predictors for the detection of phase transitions, Phys. Rev. X 12, 031044 (2022)

2022
[31]

J. Wang, C. Wang, H. Wang, and H.-Z. Lu, Unsupervised learning of topological phase transitions using the Calinski-Harabaz index, Phys. Rev. Res. 3, 013074 (2021)

2021
[32]

J. F. Rodriguez-Nieva and M. S. Scheurer, Identifying topological order through unsupervised machine learning, Nat. Phys. 15, 790 (2019)

2019
[33]

Y. Che, C. Gneiting, T. Liu, and F. Nori, Topological quantum phase transitions retrieved through unsupervised machine learning, Phys. Rev. B 102, 134213 (2020)

2020
[34]

Long and B

Y. Long and B. Zhang, Unsupervised data-driven classification of topological gapped systems with symmetries, Phys. Rev. Lett. 130, 036601 (2023)

2023
[35]

Yu and D.-L

L.-W. Yu and D.-L. Deng, Unsupervised learning of non-Hermitian topological phases, Phys. Rev. Lett. 126, 240402 (2021)

2021
[36]

S. Park, Y. Hwang, and B.-J. Yang, Unsupervised learning of topological phase diagram using topological data analysis, Phys. Rev. B 105, 195115 (2022)

2022
[37]

M. S. Scheurer and R.-J. Slager, Unsupervised machine learning and band topology, Phys. Rev. Lett. 124, 226401 (2020)

2020
[38]

Y. Long, J. Ren, and H. Chen, Unsupervised manifold clustering of topological phononics, Phys. Rev. Lett. 124, 185501 (2020)

2020
[39]

Tsai, K.-F

Y.-H. Tsai, K.-F. Chiu, Y.-C. Lai, and D.-W. Wang, Deep learning of topological phase transitions from entanglement aspects: An unsupervised way, Phys. Rev. B 104, 165108 (2021)

2021
[40]

Käming, A

N. Käming, A. Dawid, K. Kottmann, M. Lewenstein, K. Sengstock, A. Dauphin, and C. Weitenberg, Unsupervised machine learning of topological phase transitions from experimental data, Mach. Learn.: Sci. Technol. 2, 035037 (2021)

2021
[41]

Y. Long, H. Xue, and B. Zhang, Unsupervised learning of topological non-Abelian braiding in non-Hermitian bands, Nat. Mach. Intell. 6, 904 (2024)

2024
[42]

J. H. Bastek and D. M. Kochmann, Inverse design of nonlinear mechanical metamaterials via video denoising diffusion models, Nat. Mach. Intell. 5, 1466 (2023)

2023
[43]

Q. Cao, X. Li, D. Wang, C. Ma, Y. Chen, and X. Yang, Latent knowledge-guided video diffusion for scientific phenomena generation from a single initial frame, in Proceedings of the AAAI Conference on Artificial Intelligence 40, 2625–2633 (2026)

2026
[44]

Jiang and J

J. Jiang and J. A. Fan, Multiobjective and categorical global optimization of photonic structures based on ResNet generative neural networks, Nanophotonics 10, 361 (2020)

2020
[45]

Mondal, T

S. Mondal, T. Park, S. Biswas, A. X. Wang, and W. Cai, MxDiffusion: A physics-aware Maxwell's law-guided diffusion model strategy for inverse photonic metasurface design, arXiv:2602.16851

work page arXiv
[46]

Delchevalerie, N

V. Delchevalerie, N. Roy, A. Bougaham, A. Mayer, B. Frénay, and M. Lobet, Towards photonic band diagram generation with transformer-latent diffusion models, arXiv:2510.01749

work page arXiv
[47]

X. Chen, J. Sun, X. Wang, M. Chen, Q. Lin, M. Xia, and X. Zhou, Composable generation strategy framework enabled bidirectional design on topological circuits, Phys. Rev. B 110, 134108 (2024)

2024
[48]

Qwen Team, Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvement, arXiv:2409.12122

work page internal anchor Pith review arXiv
[49]

Align anything: Training all-modality models to follow instructions with language feedback,

J. Ji et al., Align Anything: Training All-Modality Models to Follow Instructions with Language Feedback, arXiv:2412.15838

work page arXiv
[50]

Yokomizo and S

K. Yokomizo and S. Murakami, Non-Bloch bands in two-dimensional non-Hermitian systems, Phys. Rev. B 107, 195112 (2023)

2023
[51]

Gururangan, A

S. Gururangan, A. Marasović, S. Swayamdipta, K. Lo, I. Beltagy, D. Downey, and N. A. Smith, Don't stop pretraining: Adapt language models to domains and tasks, in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (Association for Computational Linguistics, Online, 2020), pp. 8342–8360

2020
[52]

X. Wang, J. Wei, D. Schuurmans, Q. Le, E. Chi, S. Narang, A. Chowdhery, and D. Zhou, Self-consistency improves chain of thought reasoning in language models, arXiv:2203.11171

work page internal anchor Pith review Pith/arXiv arXiv
[53]

X. Chen, J. Sun, and X. Wang, Deep learning for the design of non-Hermitian topolectrical circuits, Phys. Rev. B 109, 094103 (2024)

2024
[54]

OpenAI, GPT-4 technical report, arXiv:2303.08774

work page internal anchor Pith review Pith/arXiv arXiv
[55]

Qwen Team, Qwen3 Technical Report, arXiv:2505.09388

work page internal anchor Pith review Pith/arXiv arXiv
[56]

DeepSeek-AI, DeepSeek LLM: Scaling open-source language models with longtermism, arXiv:2401.02954

work page internal anchor Pith review arXiv
[57]

Rombach, A

R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, High-resolution image synthesis with latent diffusion models, in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022), pp. 10684–10695

2022
[58]

Spitznagel, J

M. Spitznagel, J. Vaillant, and J. Keuper, PhysicsGen: Can generative models learn from images to predict complex physical relations?, in Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) (2025), pp. 11125–11134

2025
[59]

Fukui, Y

T. Fukui, Y. Hatsugai, and H. Suzuki, Chern numbers in discretized Brillouin zone: Efficient method of computing (spin) Hall conductances, J. Phys. Soc. Jpn. 74, 1674 (2005)

2005
[60]

Bowman, L

S. Bowman, L. Vilnis, O. Vinyals, A. Dai, R. Jozefowicz, and S. Bengio, Generating sentences from a continuous space, in Proceedings of the 20th SIGNLL Conference on Computational Natural Language Learning (Association for Computational Linguistics, Berlin, Germany, 2016), pp. 10–21

2016
[61]

E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, and W. Chen, LoRA: Low-rank adaptation of large language models, in International Conference on Learning Representations (2022)

2022
[62]

Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution

P. Wang et al., Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution, arXiv:2409.12191

work page internal anchor Pith review Pith/arXiv arXiv
[63]

Gemma Team et al., Gemma: Open Models Based on Gemini Research and Technology, arXiv:2403.08295

work page internal anchor Pith review Pith/arXiv arXiv
[64]

Llama 2: Open Foundation and Fine-Tuned Chat Models

H. Touvron et al., Llama 2: Open foundation and fine-tuned chat models, arXiv:2307.09288

work page internal anchor Pith review Pith/arXiv arXiv
[65]

Chameleon Team, Chameleon: Mixed-modal early-fusion foundation models, arXiv:2405.09818

work page internal anchor Pith review arXiv
[66]

K. Ruan, Y. Xu, Z.-F. Gao, Y. Liu, Y. Guo, J.-R. Wen, and H. Sun, Discovering physical laws with parallel symbolic enumeration, Nat. Comput. Sci. 6, 53 (2026)

2026
[67]

S. M. Udrescu and M. Tegmark, AI Feynman: A physics-inspired method for symbolic regression, Sci. Adv. 6, eaay2631 (2020)

2020
[68]

Makke and S

N. Makke and S. Chawla, Inferring interpretable models of fragmentation functions using symbolic regression, Mach. Learn.: Sci. Technol. 6, 025003 (2025)

2025
[69]

R. Wang, B. Wang, K. Li, Y. Zhang, and J. Cheng, DRSR: LLM-based scientific equation discovery with dual reasoning from data and experience, arXiv:2506.04282

work page arXiv
[70]

Llm-sr: Scientific equation discovery via programming with large language models

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:2404.18400

work page arXiv
[71]

Biggio, T

L. Biggio, T. Bendinelli, A. Neitz, A. Lucchi, and G. Parascandolo, Neural symbolic regression that scales, in Proceedings of the 38th International Conference on Machine Learning (PMLR, 2021), pp. 936–945

2021
[72]

Interpretable Machine Learning for Science with PySR and SymbolicRegression.jl

M. Cranmer, Interpretable machine learning for science with PySR and SymbolicRegression.jl, arXiv:2305.01582

work page internal anchor Pith review arXiv
[73]

J. Meng, L. Sun, X. Wang, D. Zhu and X. Zhou, Identifying topological invariants of non-Hermitian systems via domain-adaptive multimodal model for mathematics, gitHub repository (2026), https://github.com/Majic-666/Domain-Adaptive-Multimodal-Model-for-Mathematics. APPENDIX A: THE PREDICTED GBZ CURVES BY STABLE DIFFUSION Fig. 5. (a) Schematic illustration ...

2026