15 papers
MOT-SR: Multi-Objective Tool-Augmented Scientific Equation Discovery with Large Language Models
Boxiao Wang, Runxiang Wang, Kai Li +4
Symbolic Regression (SR) aims to discover analytical equations from observational data and plays a central role in scientific modeling. While recent Large Language Model (LLM) base…
AutoPref: Automatic Discovery of Task-Specific Preference Objectives for Neural Combinatorial Optimization
Shengda Gu, Kai Li, Xinyi Ke +3
AutoPref uses a large language model to automatically discover and compose pairwise loss and weighting programs that define preference objectives for neural combinatorial optimizat…
EvoPINN: Agentic Discovery of Executable Algorithms for Physics-Informed Neural Networks
Peng Yin, Kai Li, Yifan Zhang +1
EvoPINN is an agentic framework that uses a large language model to automatically generate and verify executable algorithms for physics-informed neural networks, improving the accu…
RefineEvo: Planning-Guided Heuristic Evolution with Bidirectional Experience
Yang Wu, Junran Pan, Yifan Zhang +3
RefineEvo is an evolutionary framework that uses a planner to choose operators and a reflector to store positive and negative experiences, turning automatic heuristic design into a…
When Good Equations Get Bad Scores: Improving Symbolic Regression Through Better Parameter Optimization
Boxiao Wang, Kai Li, Zhiwei Chen +5
Symbolic Regression (SR) plays a central role in scientific knowledge discovery by distilling mathematical equations from observational data. Most existing SR methods function with…
Reflex: Reinforcement Learning with Reflection Symmetry Exploitation in State-Based Continuous Control
Shuai Zhen, Yifan Zhang, Yuling Wang +1
Reinforcement learning has long struggled with poor sample efficiency. One promising approach to mitigate this problem is leveraging group-invariant Markov Decision Processes (-…