SymTorch allows you to approximate the behaviour of components within deep learning models with symbolic equations using PySR.
SymTorch is available on PyPI as torch-symbolic:
pip install torch-symbolicOur paper is available on arXiv. You can also visit the accompanying website for more information.
Full documentation is available at ReadTheDocs.
If you use SymTorch in your research, please cite:
@misc{symtorch2026,
title = {SymTorch: A Framework for Symbolic Distillation of Deep Neural Networks},
author = {Tan, Elizabeth S.Z. and Soubki, Adil and Cranmer, Miles},
year = {2026},
eprint = {2602.21307},
archivePrefix= {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2602.21307}
}MIT License