{"total":56,"items":[{"citing_arxiv_id":"2605.22498","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"The Neural Compiler: Program-to-Network Translation for Hybrid Scientific Machine Learning","primary_cat":"cs.LG","submitted_at":"2026-05-21T13:49:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"The Neural Compiler converts symbolic programs into exact differentiable PyTorch modules for hybrid scientific machine learning, enabling precise encoding of known physics with few trainable parameters.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.22374","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Guiding Multi-Objective Genetic Programming with Description Length Improves Symbolic Regression Solutions","primary_cat":"cs.NE","submitted_at":"2026-05-21T12:07:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Post-selection with DL or FBF after multi-objective GP search improves test-set performance over AIC/BIC baselines on noisy synthetic and real regression tasks, while using DL directly as fitness often causes premature convergence to overly simple models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.22330","ref_index":41,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Symbolic Classification-Enabled LHC Limits Online BSM Global Fits","primary_cat":"hep-ph","submitted_at":"2026-05-21T11:19:27+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Symbolic regression produces an approximate classifier for LHC exclusion limits that enables their direct inclusion during pMSSM global fits.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.21813","ref_index":2,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Symbolic Density Estimation for Discrete Distributions","primary_cat":"cs.LG","submitted_at":"2026-05-20T23:22:21+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"SDE recovers closed-form PMFs for discrete distributions via evolutionary search guided by domain priors, recovering all benchmark families with accurate parameters and improving mixture fits on real data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.19379","ref_index":23,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Graph-based automated discovery of concise soil hydraulic functions from data: beyond the Mualem - van Genuchten model","primary_cat":"physics.flu-dyn","submitted_at":"2026-05-19T05:25:53+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A graph-based automated model discovery framework identifies new concise soil hydraulic functions from data that outperform the Mualem-van Genuchten model across 249 soil samples.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.17790","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"STRIDE: A Self-Reflective Agent Framework for Reliable Automatic Equation Discovery","primary_cat":"cs.AI","submitted_at":"2026-05-18T03:14:32+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"STRIDE is a self-reflective agent framework that improves accuracy, OOD robustness, and structural recovery in LLM-based symbolic regression by integrating generation, evaluation, repair, and diversity-preserving memory.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.18883","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Prediction Is Not Physics: Learning and Evaluating Conserved Quantities in Neural Simulators","primary_cat":"cs.LG","submitted_at":"2026-05-16T05:23:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Conservation Discovery Networks recover analytical energy with R² ≥ 0.996 in Hamiltonian systems using temporal consistency and λ_align=0.2, but collapse without alignment and show mixed noise robustness.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.16724","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Discovering interpretable low-dimensional dynamics using maximum entropy","primary_cat":"q-bio.QM","submitted_at":"2026-05-16T00:18:21+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Edwin integrates dynamic maximum entropy dimensionality reduction with symbolic regression to recover physically interpretable low-dimensional dynamics from high-dimensional observations that generalize to unseen conditions.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.15809","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Diversified Residual Symbolic Regression","primary_cat":"cs.NE","submitted_at":"2026-05-15T10:04:03+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"DRSR uses Quality-Diversity to produce diverse symbolic regression expressions differing in residual distributions, enabling post-search selection on synthetic and astronomical data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.15197","ref_index":233,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Primordial Black Hole from Tensor-induced Density Fluctuation: First-order Phase Transitions and Domain Walls","primary_cat":"astro-ph.CO","submitted_at":"2026-05-14T17:59:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Tensor perturbations from first-order phase transitions and domain wall annihilation induce curvature fluctuations at second order that form primordial black holes, allowing asteroid-mass PBHs to comprise all dark matter for specific parameter ranges with associated gravitational wave peaks in LISA,","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Cranmer, arXiv preprint arXiv:2305.01582 (2023). [230] D. I. Dunsky and M. Kongsore, JHEP06, 198 (2024), arXiv:2402.03426 [hep-ph]. [231] R. Z. Ferreira, A. Notari, O. Pujol` as, and F. Rompineve, JCAP06, 020 (2024), arXiv:2401.14331 [astro-ph.CO]. [232] B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. Lett.116, 061102 (2016), arXiv:1602.03837 [gr-qc]. [233] B. P. Abbott et al. (LIGO Scientific, Virgo), Phys. Rev. Lett.116, 241103 (2016), arXiv:1606.04855 [gr-qc]. [234] B. P. Abbott et al. (LIGO Scientific, VIRGO), Phys. Rev. Lett.118, 221101 (2017), [Erratum: Phys.Rev.Lett. 121, 129901 (2018)], arXiv:1706.01812 [gr-qc]. [235] B. . P. . Abbott et al. (LIGO Scientific, Virgo), Astrophys. J. Lett.851, L35 (2017), arXiv:1711."},{"citing_arxiv_id":"2605.12704","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"FePySR: A Neural Feature Extraction Framework for Efficient and Scalable Symbolic Regression","primary_cat":"cs.SC","submitted_at":"2026-05-12T20:04:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"FePySR uses a neural network to pre-extract valid features before PySR search, recovering more equations than baselines on benchmarks and identifying governing ODEs in 24 of 100 biological cases where PySR finds none.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.11280","ref_index":87,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Discovery of Interpretable Surrogates via Agentic AI: Application to Gravitational Waves","primary_cat":"gr-qc","submitted_at":"2026-05-11T22:09:34+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"GWAgent agentic workflow produces analytic surrogates for eccentric BBH waveforms with 6.9e-4 median mismatch and 8.4x speedup, outperforming baselines, and infers eccentricity for GW200129.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"the surrogate can support end-to-end astrophysical inference in addition to rapid waveform generation. Comparison with alternative methods To benchmark performance, we train conventional machine- learning and symbolic-regression algorithms on the same resid- ual data. We employ state-of-the-art symbolic-regression pack- ages such aspySR[86], gplearn [87], Operon [88], and AI- Feynman [89] in a one-shot mode. Figure 5 summarizes the resulting accuracy-cost trade-offs. Most baselines have MM LIGO >10 −3, while the agentic surrogate lies on the best Pareto frontier. The cost points pile up near∼12 ms because, beyond this point, waveform evaluation costs are dominated by phase integration rather than residual evaluation itself."},{"citing_arxiv_id":"2605.10687","ref_index":12,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"The finite expression method for turbulent dynamics with high-order moment recovery","primary_cat":"cs.LG","submitted_at":"2026-05-11T15:00:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A two-stage symbolic regression plus generative model framework recovers governing interaction terms and forcing in stochastic triad models while accurately predicting statistical moments up to order five.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"for adaptive learning of stochastic partial differential equation solutions.arXiv preprint arXiv:2508.06834, 2025. [10] Diederik P Kingma and Max Welling. Auto-encoding variational bayes.arXiv preprint arXiv:1312.6114, 2013. [11] Jasen Lai, Senwei Liang, and Chunmei Wang. H-fex: A symbolic learning method for hamiltonian systems.arXiv preprint arXiv:2506.20607, 2025. [12] Brendan Lenfesty, Saugat Bhattacharyya, and KongFatt Wong-Lin. Uncovering dynamical equations of stochastic decision models using data-driven sindy algorithm.Neural Computation, 37(3):569-587, 2025. [13] SENWEI LIANG, Chunmei Wang, and XINGJIAN XU. Identifying stochastic dynamics via finite expression methods.Available at SSRN 5327407. [14] Senwei Liang and Haizhao Yang."},{"citing_arxiv_id":"2605.10685","ref_index":8,"ref_count":3,"confidence":0.98,"is_internal_anchor":true,"paper_title":"GESR: A Genetic Programming-Based Symbolic Regression Method with Gene Editing","primary_cat":"cs.AI","submitted_at":"2026-05-11T15:00:22+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"GESR uses two BERT models to intelligently direct mutations and crossovers inside genetic programming, yielding higher efficiency and competitive accuracy on symbolic regression benchmarks.","context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"(GP) [5-7] extends GA to symbolic regression by representing mathematical expres- sions as tree structures and evolving them through crossover and mutation. Due to its flexible representation, GP remains one of the most widely used paradigms in sym- bolic regression. However, standard GP often suffers from inefficient random mutation, slow convergence, and expression bloat. PySR [8] improves GP by integrating constant optimization, evolutionary search, and Pareto-based complexity control, enabling the discovery of compact and interpretable symbolic expressions. Reinforcement Learning-Based Methods.Deep Symbolic Regression (DSR) [9] formulates symbolic regression as a reinforcement learning problem, where a recurrent neural network (RNN) acts as a policy to sequentially generate expression"},{"citing_arxiv_id":"2605.09696","ref_index":43,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Discovery of Nonlinear Dynamics with Automated Basis Function Generation","primary_cat":"cs.LG","submitted_at":"2026-05-10T18:30:17+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AutoSINDy automatically builds a tailored basis library from PySR symbolic regression and applies SINDy to recover ground-truth nonlinear dynamics with 92.8% success under noise.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.07323","ref_index":2,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Discovering Ordinary Differential Equations with LLM-Based Qualitative and Quantitative Evaluation","primary_cat":"cs.AI","submitted_at":"2026-05-08T06:29:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DoLQ employs a sampler agent, parameter optimizer, and LLM-based scientist agent to iteratively propose, refine, and evaluate ODE candidates, yielding higher success rates and better symbolic term recovery than prior symbolic regression methods on multi-dimensional benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.06123","ref_index":65,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Back to the Beginning of Heuristic Design: Bridging Code and Knowledge with LLMs","primary_cat":"cs.AI","submitted_at":"2026-05-07T12:30:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A knowledge-first approach to LLM-driven automatic heuristic design in combinatorial optimization yields better discovery efficiency, transfer, and generalization than code-centric baselines by formalizing a distortion-compression trade-off.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.03101","ref_index":17,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Programmatic Context Augmentation for LLM-based Symbolic Regression","primary_cat":"cs.AI","submitted_at":"2026-05-04T19:34:03+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Programmatic context augmentation lets LLM-based symbolic regression perform code-driven data analysis during search, yielding superior efficiency and accuracy over baselines on LLM-SRBench.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.01072","ref_index":33,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Reconstructing conformal field theoretical compositions with Transformers","primary_cat":"hep-th","submitted_at":"2026-05-01T20:09:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Transformers reconstruct the constituent RCFTs in tensor-product theories from low-energy spectra, reaching 98% accuracy on WZW models and generalizing to larger central charges with few out-of-domain examples.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"In addition, [30] and [31] both applied ML techniques such as deep neural networks and Reinforcement Learning (RL) to study other properties of 2Dor 3DCFTs such as correlation functions and OPE coefficients. The investigation of energy spectra of 2D RCFTs in symbolic forms was performed in [32] with a symbolic regression package in python called pySR [33]. More recently, Benjamin et al [34] investigated the landscape of two-dimensional CFTs by searching for numerical solutions to the modular bootstrap equation with stochastic sampling optimization. 2. The Basics of Conformal Field Theory Similar to statistical physics, the spectra of CFTs are embedded in the partition functions as conformal \"tower\" of states."},{"citing_arxiv_id":"2605.08130","ref_index":4,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Additive Atomic Forests for Symbolic Function and Antiderivative Discovery","primary_cat":"cs.LG","submitted_at":"2026-05-01T04:27:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"UNKNOWN","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A derivative algebra with EML and SOL primitives plus additive atomic forests enables simultaneous symbolic recovery of functions and antiderivatives from data, matching or exceeding XGBoost on 13 of 17 benchmarks with interpretable formulas.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.27297","ref_index":20,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Machine Collective Intelligence for Explainable Scientific Discovery","primary_cat":"cs.AI","submitted_at":"2026-04-30T01:15:54+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Machine collective intelligence uses coordinated AI agents to evolve symbolic hypotheses and recover governing equations from observations in deterministic, stochastic, and uncharacterized systems, achieving up to six orders of magnitude better extrapolation than neural networks with 5-40 parameters","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.20941","ref_index":11,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Interpretable Analytic Formulae for GWTC-4 Binary Black Hole Population Properties via Symbolic Regression","primary_cat":"astro-ph.CO","submitted_at":"2026-04-22T15:04:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Symbolic regression on GWTC-4 posteriors yields closed-form analytic formulae for merger-rate evolution, effective-spin dependencies on mass ratio and redshift, and conditional mass-ratio distributions at specific primary mass peaks.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"probability of monotonic broadening equal to unity (Pr[broadens] = 1.0). TheSplinemodel permits more complex dispersion dynamics, but SR firmly captures an initial low-redshift broadening phase. The optimal symbolic surrogate mathematically encodes this transition using an absolute-value term (|z−1|) to serve as a functional hinge atz= 1: logσ χeff (z)≃0.15|z−1| 1.61z2 −0.23z−1.69 \u0001 −1.45(z−1) 2 −1.50 (11) This expression elegantly distills the Spline model's complex median behavior. By expanding the polynomial for local redshifts (z <1), the equation yields a steep, continuous broadening within the low-zwindow (average gradient≈+2.60). As evolution crossesz= 1, the absolute-value hinge smoothly inverts the polynomial contri- bution, transitioning the distribution into a high-redshift plateau and gradual narrowing phase (average gradient"},{"citing_arxiv_id":"2604.18548","ref_index":32,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Physics-Informed Neural Networks for Biological $2\\mathrm{D}{+}t$ Reaction-Diffusion Systems","primary_cat":"cs.LG","submitted_at":"2026-04-20T17:40:45+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"BINNs are extended to 2D+t systems and combined with symbolic regression to recover reaction-diffusion models of lung cancer cell dynamics from time-lapse microscopy data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.18414","ref_index":28,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Balance-Guided Sparse Identification of Multiscale Nonlinear PDEs with Small-coefficient Terms","primary_cat":"cs.LG","submitted_at":"2026-04-20T15:31:47+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"BG-SINDy reformulates l0-constrained regression as term-level l2,0 regularization and uses progressive pruning guided by balance contributions to recover small-coefficient terms in multiscale PDEs.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"the AI Feynman algorithm, which integrates neural-network guidance with brute-force symbolic search [ 25]. This laid the foundation for advances in evolutionary computation, including symbolic genetic algorithms [ 26] and ef- ficient modern implementations such as PySR [ 27], which excel in chaotic and nonlinear dynamical systems. Kolmogorov-Arnold networks (KANs) [ 28, 29] have further bridged neural networks and symbolic regression by learning spline-based univariate functions that can be directly translated into ex- plicit symbolic expressions. More recently, large-language-model-assisted approaches have emerged that dynamically update candidate libraries or generate novel expressions [ 30]. At the current frontier, generative trans-"},{"citing_arxiv_id":"2604.17470","ref_index":40,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Machine Learning Hamiltonian Dynamical Systems with Sparse and Noisy Data","primary_cat":"cs.LG","submitted_at":"2026-04-19T14:51:43+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ASRNNs recover Hamiltonian dynamics and symbolic equations from trajectories with only two irregularly spaced noisy points by preserving symplectic structure without derivative estimation.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.16842","ref_index":80,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Singularity Formation: Synergy in Theoretical, Numerical and Machine Learning Approaches","primary_cat":"math.NA","submitted_at":"2026-04-18T05:24:35+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"The work introduces a modulation-based analytical method for singularity proofs in singular PDEs and refines ML techniques like PINNs and KANs to identify blowup solutions, with application to the open 3D Keller-Segel problem.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"differentiableway[134,111,304,381]orsearchedinadiscreteway[29]. Activation function are parametrized as polynomials [134], splines [111, 31, 10], sigmoid linear unit [304], or neural networks [381]. KANs use B-splines to parametrize their activation functions. 64 Symbolic Regression.There are many off-the-shelf symbolic regression methods based on genetic algorithms (Eureka [96], GPLearn [135], PySR [80]), neural- networkbasedmethods(EQL[248], OccamNet[97]), physics-inspiredmethod(AI Feynman [346, 347]), and reinforcement learning-based methods [268]. KANs are mostsimilartoneuralnetwork-basedmethods,butdifferfrompreviousworksinthat our activation functions are continuously learned before symbolic snapping rather than manually fixed [96, 97]. Physics-InformedNeuralNetworks(PINNs)andPhysics-InformedNeuralOp-"},{"citing_arxiv_id":"2604.16232","ref_index":12,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Neuro-Symbolic ODE Discovery with Latent Grammar Flow","primary_cat":"cs.LG","submitted_at":"2026-04-17T16:46:23+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Latent Grammar Flow discovers ODEs by placing grammar-based equation representations in a discrete latent space, using a behavioral loss to cluster similar equations, and sampling via a discrete flow model guided by data fit and constraints.","context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"the outer optimisation loop (including the initialisation) with a population size ofn pop = 100. As this benchmark only contains first-order explicit ODEs, neither the order nor the stability predictors (i.e. guidance) was applied. Hence, the generative process was guided by the optimisation objective. LGF is compared with the transformer-based generative method ODEFormer [27], the evolutionary algorithm-based search method PySR [12], a grammar-based search method ProGED [13, 14] and a hybrid grammar-based method GODE [23]. A description of these methods can be found in Section A. The resulting mean relative L2 errors are listed in Table 2, including the mean and standard deviation (std) of the complexity. Figure 2 depicts the distribution of the relative L2 errors in violin plots."},{"citing_arxiv_id":"2604.16015","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Discovering quantum phenomena with Interpretable Machine Learning","primary_cat":"quant-ph","submitted_at":"2026-04-17T12:44:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Variational autoencoders combined with symbolic regression extract physically meaningful representations and order parameters from raw quantum measurement data, revealing new phenomena such as corner-ordering in Rydberg arrays.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.13423","ref_index":28,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Into the Gompverse: A robust Gompertzian reionization model for CMB analyses","primary_cat":"astro-ph.CO","submitted_at":"2026-04-15T02:49:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A Gompertzian reionization model with three nuisance parameters demotes optical depth to a derived quantity, reducing its uncertainty by a factor of three and revealing potential neutrino mass tension in CMB analyses.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"4) and our 1024xHI(z) samples we can establish the mapping between cosmological parameters and reionization through the pivot and rescaling of each simulation. In total, we have 4 + 2×512 parameters to determine the joint fit. The first four parameters correspond to the coefficientsc m3 while the fittedαandβare then used as our targets for symbolic regression. We use symbolic regression [27] - via thePySRpackage [28] - to search a vast function space for the expressions of lnαandβas functions of cosmology and astrophysical parameters. Our search space encompasses the set of expressions composed of the operators +,−,·,/, power, exp, and ln. We point readers interested in the specific details of the implementation, including the use of the Pareto front for model selection, to the methodology described in [4]"},{"citing_arxiv_id":"2604.11249","ref_index":34,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"First observational constraints on cosmic backreaction over an extended redshift range","primary_cat":"astro-ph.CO","submitted_at":"2026-04-13T09:59:16+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"First direct constraints on total cosmic backreaction over a significant redshift range are consistent with vanishing backreaction within 1 sigma but are too weak to exclude meaningful backreaction.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.06998","ref_index":72,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Identifying Topological Invariants of Non-Hermitian Systems via Domain-Adaptive Multimodal Model for Mathematics","primary_cat":"cond-mat.other","submitted_at":"2026-04-08T12:15:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"A domain-adaptive multimodal model with a mathematics LLM backbone identifies topological invariants of non-Hermitian systems from eigenvalues and eigenvectors in momentum space.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.05822","ref_index":47,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Model-independent constraints on generalized FLRW consistency relations with bootstrap-based symbolic regression","primary_cat":"astro-ph.CO","submitted_at":"2026-04-07T12:54:53+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Bootstrap-based symbolic regression on supernova and BAO data finds mild 2-4 sigma deviations from FLRW consistency relations, which if real would rule out most FLRW-based solutions to cosmological tensions.","context_count":1,"top_context_role":"method","top_context_polarity":"use_method","context_text":"retaining or discarding symbolic expressions when analyzing both real data and a new, substantially larger set of mock data. Having calibrated the selection criteria using the initial five exploratory bootstrap samples, we move on to perform an actual robustness test of the method. To do this, we generate 200 new bootstrap sam- ples based on a ΛCDM model withH 0 = 72km/s/Mpc17. According to [47], 50-200 samples is sufficient to obtain accurate estimates of the median and uncertainty in most situations. Using these new bootstrap samples, we then follow the procedure described above and summarized here: dA Procedure: 1. Run cp3-bench for each bootstrap sample and man- ually inspect the mean-square-error andd ′ A, d′′ A of each resulting symbolic expression."},{"citing_arxiv_id":"2603.25111","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"SEVerA: Verified Synthesis of Self-Evolving Agents","primary_cat":"cs.LG","submitted_at":"2026-03-26T07:32:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"SEVerA uses Formally Guarded Generative Models and a three-stage Search-Verification-Learning process to synthesize self-evolving agents that satisfy hard formal constraints while improving task performance.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2603.20910","ref_index":11,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"LLM-ODE: Data-driven Discovery of Dynamical Systems with Large Language Models","primary_cat":"cs.LG","submitted_at":"2026-03-21T18:46:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"LLM-ODE integrates large language models into genetic programming to guide symbolic search for governing equations of dynamical systems, outperforming classical GP on 91 test cases in efficiency and solution quality.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.01231","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Experimental Design for Missing Physics","primary_cat":"stat.ML","submitted_at":"2026-03-21T12:26:21+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"A sequential experimental design technique discriminates between model structures from symbolic regression to discover missing physics in process systems such as bioreactors.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2603.15250","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"In-Context Symbolic Regression for Robustness-Improved Kolmogorov-Arnold Networks","primary_cat":"cs.LG","submitted_at":"2026-03-16T13:21:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"In-context symbolic regression methods improve robustness of symbolic formula recovery from KANs, cutting median OFAT test MSE by up to 99.8 percent across hyperparameter sweeps.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.15603","ref_index":27,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Symbolic recovery of PDEs from measurement data","primary_cat":"cs.LG","submitted_at":"2026-02-17T14:20:36+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Symbolic rational-function networks recover an admissible PDE from noiseless complete measurements and select the regularization-minimizing parameterization within the architecture.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2601.22324","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Automatic Construction of Clinical Scoring Systems with LLM Agents","primary_cat":"cs.LG","submitted_at":"2026-01-29T21:11:06+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AgentScore uses LLM agents for semantically guided search over clinical scoring rules combined with data-driven verification, outperforming prior score generation methods on eight tasks and established guidelines on two externally validated tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2601.16282","ref_index":2,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Generating Literature-Driven Scientific Theories at Scale","primary_cat":"cs.CL","submitted_at":"2026-01-22T19:27:49+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Literature-grounded LLM synthesis of theories from 13.7k papers yields 2.9k theories that better match evidence and predict future results from 4.6k subsequent papers than parametric baselines.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2512.15920","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Introduction to Symbolic Regression in the Physical Sciences","primary_cat":"cs.LG","submitted_at":"2025-12-17T19:32:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":2.0,"formal_verification":"none","one_line_summary":"Symbolic regression provides an interpretable way to extract mathematical relationships from data for scientific discovery and surrogate modeling in the physical sciences.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2512.15567","ref_index":54,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Evaluating Large Language Models in Scientific Discovery","primary_cat":"cs.AI","submitted_at":"2025-12-17T16:20:03+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"The SDE benchmark shows LLMs lag on scientific discovery tasks relative to general science tests, with diminishing scaling returns and shared weaknesses across models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2512.06044","ref_index":80,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"HoloNet: Toward a Unified Einstein-Maxwell-Dilaton Framework of QCD","primary_cat":"hep-lat","submitted_at":"2025-12-05T05:23:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A neural network learns holographic bulk functions from lattice QCD data at zero chemical potential and embeds them into an EMD model to describe finite-density QCD and locate the critical end point.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2511.09644","ref_index":27,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"statmorph-lsst: Quantifying and correcting morphological biases in galaxy surveys","primary_cat":"astro-ph.GA","submitted_at":"2025-11-12T19:00:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Morphological metrics in galaxy images suffer systematic biases from resolution, depth, and noise that can be quantified and corrected empirically, with new metrics proposed to reduce those effects.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2511.07686","ref_index":34,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Kolmogorov-Arnold Chemical Reaction Neural Networks for learning pressure-dependent kinetic rate laws","primary_cat":"physics.chem-ph","submitted_at":"2025-11-10T23:08:54+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"KA-CRNNs learn pressure-dependent and collider-specific kinetic rate laws from data using Kolmogorov-Arnold activations inside a CRNN framework, outperforming interpolative methods by 2.88x in MSE on two proof-of-concept reactions.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2510.25781","ref_index":17,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"A Practitioner's Guide to Kolmogorov-Arnold Networks","primary_cat":"cs.LG","submitted_at":"2025-10-28T03:03:44+00:00","verdict":"ACCEPT","verdict_confidence":"LOW","novelty_score":3.0,"formal_verification":"none","one_line_summary":"A systematic review of Kolmogorov-Arnold Networks that maps their relation to Kolmogorov superposition theory, MLPs, and kernels, examines basis-function design choices, summarizes performance advances, and supplies a practitioner's selection guide plus open challenges.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Collectively, these developments establish MLP-PINNs as a mature reference standard in scientific machine learning, many of which, as we show in Se ction 4.1, have directly inspired KAN development. Despite their flexibility, MLPs face well-documented limitations. Fixed a ctivation functions restrict adapt- ability [ 16]. Network behavior can be difficult to interpret [ 17] and to attribute causally [ 18]. Achieving high accuracy often entails large parameter counts, which can hinder e fficient updates [ 19, 20]. Generalization robustness can degrade in challenging regimes [ 21, 269]. Optimization itself can be fragile or stiff, depending on the task and scaling [ 22]. MLPs also exhibit spectral bias: a tendency to learn low frequencie s faster than"},{"citing_arxiv_id":"2510.06390","ref_index":20,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Proper time expansions and glasma dynamics","primary_cat":"nucl-th","submitted_at":"2025-10-07T19:13:28+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"The authors test methods that extend the reliable reach of proper time expansions for glasma dynamics from roughly 0.05 fm/c to about 0.08 fm/c.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2508.20266","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Using Symbolic Regression to Emulate the Radial Fourier Transform of the S\\'ersic profile for Fast, Accurate and Differentiable Galaxy Profile Fitting","primary_cat":"astro-ph.IM","submitted_at":"2025-08-27T20:47:47+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Symbolic regression yields an emulator for the radial Fourier transform of the Sérsic profile that enables 2.5 times faster galaxy profile fitting with minimal accuracy loss.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2507.17777","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"ASP-Assisted Symbolic Regression: Uncovering Hidden Physics in Fluid Mechanics","primary_cat":"cs.AI","submitted_at":"2025-07-22T15:16:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A hybrid symbolic regression and answer set programming framework derives compact, physically plausible equations for velocity and pressure in 3D laminar channel flow from simulation data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2507.10462","ref_index":52,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"What is the diatomic molecule with the largest dipole moment?","primary_cat":"physics.atom-ph","submitted_at":"2025-07-14T16:39:45+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"A machine learning model based on atomic properties predicts diatomic dipole moments, screens the periodic table for the largest values, and condenses into an analytical expression.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2506.13131","ref_index":21,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"AlphaEvolve: A coding agent for scientific and algorithmic discovery","primary_cat":"cs.AI","submitted_at":"2025-06-16T06:37:18+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"AlphaEvolve is an LLM-orchestrated evolutionary coding agent that discovered a 4x4 complex matrix multiplication algorithm using 48 scalar multiplications, the first improvement over Strassen's algorithm in 56 years, plus optimizations for Google data centers and hardware.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Related work Evolutionary methods. AlphaEvolveextends a long tradition of research onevolutionary or genetic programming[54], where one repeatedly uses a set of mutation and crossover operators to evolve a pool of programs [5,51]. In particular, classical evolutionary techniques have succeeded in symbolic regression applications [66, 87], automated scientific [21] or algorithmic [16] discovery, and scheduling [118] problems. However, a challenge with these 18 AlphaEvolve: A coding agent for scientific and algorithmic discovery methods is the use of handwritten evolution operators, which can be hard to design and may fail to capture important properties of the domain. In contrast,AlphaEvolveuses LLMs to automate the construction of these operators-it leverages the LLM's world knowledge to"}],"limit":50,"offset":0}