This repo is for the paper:
Position: Reasoning is a Learnable Rule-Based Process. The 43rd International Conference on Machine Learning (ICML 2026).
View our project page here: https://jmaasch.github.io/reasoning/
reasoning_algorithm.ipynb: This notebook provides an example implementation of Algorithm 1: Valid reasoning as exact rule application. Operationalization of Definition 2.4 (reasoning) and Algorithm 1 is illustrated by implementing a trivial logical reasoning process in Python, offering a simple demonstration for how each core component can be instantiated in compilable code and how validity can be auditable.
Please cite our work using the following BibTeX.
@inproceedings{lawrence2026reasoning,
title={Position: Reasoning is a Learnable Rule-Based Process},
author={Lawrence, Rachel and Maasch, Jacqueline},
booktitle={The 43rd International Conference on Machine Learning (ICML 2026)},
year={2026}
}