RaLU

This repo is for our paper Reasoning-as-Logic-Units: Scaling Test-Time Reasoning in Large Language Models Through Logic Unit Alignment accepted by ICML'25.

Run

1. Set the url and api-key in src/bot.py
2. Run main.py --dataset <dataset> --model <model> in src.

Due to space limit, we only upload the dataset of HumanEval. All benchmarks are publicly availabile and we will also upload them after the anonymity period.

Citation

@inproceedings{li2025RaLU,
  author       = {Cheryl Li and Tianyuan Xu and Ryman Guo},
  title        = {Reasoning-as-Logic-Units: Scaling Test-Time Reasoning in Large Language Models Through Logic Unit Alignment},
  booktitle    = {International Conference on Machine Learning, {ICML} 2025, 13-19 July 2025, Vancouver, Canada},
  series       = {Proceedings of Machine Learning Research},
  publisher    = {{PMLR}},
  year         = {2025},
  url          = {[https://proceedings.mlr.press/v202/gao23f.html](https://www.arxiv.org/abs/2502.07803)},
}