MOT-SR

Code and reproducibility materials for the paper “MOT-SR: Multi-Objective Tool-Augmented Symbolic Regression via LLMs”.

• 2026-11-13 — MOT-SR is accepted to KDD 2027 as an Oral presentation 🎉
• 2026-07-31 — Preprint released on arXiv:2607.29561

• Environment: see requirements.txt or environment.yml
• Main entry: main.py
• Example script: run_motsr.sh

See the English and 中文 reproducibility guides for the environment, prompts, experiment settings, random seeds, API usage and runtime logs.

The four benchmark equation structures from MOT-SR (LLaMA-3.1) and the fitted EMRI correction are available in results/final_equations.

EMRI case generation code: FSI and PN5.

@misc{wang2026motsrmultiobjectivetoolaugmentedscientific,
      title={MOT-SR: Multi-Objective Tool-Augmented Scientific Equation Discovery with Large Language Models}, 
      author={Boxiao Wang and Runxiang Wang and Kai Li and Chongming Li and Zhiwei Chen and Yifan Zhang and Jian Cheng},
      year={2026},
      eprint={2607.29561},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2607.29561}, 
}