From fundamental physics to gravitational-wave astronomy, computational scientific discovery increasingly relies on sophisticated algorithms to analyze complex datasets, yet reliable identification of gravitational-wave signals with unknown source parameters buried in dynamic detector noise remains a formidable challenge. While existing algorithmic approaches have achieved partial success, their core limitations arise from restrictive prior assumptions: traditional methods suffer from reliance on predefined theoretical priors, while neural network approaches introduce hidden biases and lack interpretability.
This paper introduces Evolutionary Monte Carlo Tree Search (Evo-MCTS), the first integration of large language model (LLM) guidance with domain-aware physical constraints to generate interpretable solutions for automated gravitational wave detection. This framework systematically explores algorithmic solution spaces through tree-structured search enhanced by evolutionary optimization. Experimental validation demonstrates substantial performance improvements, achieving notable improvements over state-of-the-art gravitational wave detection algorithms on the MLGWSC-1 benchmark dataset and significant advantages over other LLM-based algorithm optimization frameworks.
More fundamentally, our framework establishes a transferable methodology for automated algorithmic discovery across computational science domains through systematic exploration of novel algorithmic combinations.
$ git clone https://github.com/iphysresearch/evo-mcts.git # download evo-mcts code
$ cd evo-mcts
$ python3 -m venv ./venv # [optional] create virtual environment
$ source ./venv/bin/activate # [optional] activate virtual environment
$ pip install hydra-core tqdm numpy joblib openai h5py gwpy pandas lalsuite
$ pip install "scipy==1.12.0" # dependency specifically required
$ export OPENAI_API_KEY=xxxxxxxxxx # set OpenAI API key
$ python main.py problem=gw_mlgwsc1$ export DEEPSEEK_API_KEY=xxxxxxxxxxxxxxxxxxxx
$ python main.py llm_client=deepseek llm_client.model=deepseek-chat@article{wang2025automated,
title={Automated Algorithmic Discovery for Gravitational-Wave Detection Guided by LLM-Informed Evolutionary Monte Carlo Tree Search},
author={He Wang and Liang Zeng},
year={2025},
eprint={2508.03661},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2508.03661}
}