SymBOL: A General-Purpose Symbolic Learner for Scientific Discovery Using Bayesian Optimization-Enhanced Large Language Models
- 1 November 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48 (11) , 14226-14238
Abstract
Symbolic Regression (SR) is a core challenge in both physics and artificial intelligence, aiming to identify mathematical equations from experimental data. Recent impressive advancements in Large Language Model (LLM)-based SR methods have addressed the limitations of traditional approaches in flexibly incorporating prior knowledge to enhance accuracy, but they still face challenges such as high costs and scalability with numerous variables. To overcome these issues, we introduce SymBOL. This general-purpose symbolic learning framework uses Bayesian Optimization (BO) to guide the generation of high-quality mathematical expressions from the LLM while accommodating complex tasks with a large number of interdependent variables. Experiments on benchmark datasets demonstrate that SymBOL significantly outperforms baseline methods in both accuracy and efficiency. Notably, compared to advanced LLM-based approaches, SymBOL achieves a 24.85% improvement in average accuracy while reducing computational costs by 28.73%. This advantage extends to high-dimensional SR tasks, where SymBOL substantially lowers the average error. Furthermore, when applied to real-world systems in materials science and epidemiology, SymBOL accurately recovers governing equations and provides interpretable pathways for equation discovery. These findings underscore the potential of SymBOL for advancing scientific discovery.Keywords
Funding Information
- National Natural Science Foundation of China (U22A2098)
- Deep Earth National Science and Technology Major Project (2024ZD1002903)
- Natural Science Foundation of Jilin Province (20250102211JC)
- Major Science and Technology Development Plan of Changchun (2024WX05)
This publication has 43 references indexed in Scilit:
- Learning continuous network emerging dynamics from scarce observations via data-adaptive stochastic processesScience China Information Sciences, 2024
- Interpretable scientific discovery with symbolic regression: a reviewArtificial Intelligence Review, 2024
- D-dimensional oscillators in simplicial structures: Odd and even dimensions display different synchronization scenariosChaos, Solitons & Fractals, 2021
- Networks beyond pairwise interactions: Structure and dynamicsPhysics Reports, 2020
- An interactive web-based dashboard to track COVID-19 in real timeThe Lancet Infectious Diseases, 2020
- Stress-strain data for aluminum 6061-T651 from 9 lots at 6 temperatures under uniaxial and plane strain tensionData in Brief, 2019
- Universal resilience patterns in complex networksNature, 2016
- Implementing the Nelder-Mead simplex algorithm with adaptive parametersComputational Optimization and Applications, 2010
- Excitatory and Inhibitory Interactions in Localized Populations of Model NeuronsBiophysical Journal, 1972
- Species packing and competitive equilibrium for many speciesTheoretical Population Biology, 1970