This repository contains the code used in the paper Are Language Models Efficient Reasoners? A Perspective from Logic Programming, NeurIPS 2025. Data generation is done by adapting the MathGAP backend. All code specific to this paper can be found here.
The generated datasets used for evaluation can be found here.
@inproceedings{opedal2025efficientreasoners,
title = {Are Language Models Efficient Reasoners? {A} Perspective from Logic Programming},
author = {Opedal, Andreas and Zengaffinen, Yanick and Shirakami, Haruki and Pasti, Clemente and Sachan, Mrinmaya and Saparov, Abulhair and Cotterell, Ryan and Schölkopf, Bernhard},
year = {2025},
booktitle = {The Thirty-ninth Annual Conference on Neural Information Processing Systems},
eprint = {2510.25626},
archiveprefix = {arXiv},
primaryclass = {cs.CL},
url = {https://arxiv.org/abs/2510.25626},
}