Abstract-level Deductive Reasoning for Pre-trained Language Models

Xin Wu, Yi Cai, Ho-fung Leung


Abstract
Pre-trained Language Models have been shown to be able to emulate deductive reasoning in natural language. However, PLMs are easily affected by irrelevant information (e.g., entity) in instance-level proofs when learning deductive reasoning. To address this limitation, we propose an Abstract-level Deductive Reasoner (ADR). ADR is trained to predict the abstract reasoning proof of each sample, which guides PLMs to learn general reasoning patterns rather than instance-level knowledge. Experimental results demonstrate that ADR significantly reduces the impact of PLMs learning instance-level knowledge (over 70%).
Anthology ID:
2024.lrec-main.6
Volume:
Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)
Month:
May
Year:
2024
Address:
Torino, Italia
Editors:
Nicoletta Calzolari, Min-Yen Kan, Veronique Hoste, Alessandro Lenci, Sakriani Sakti, Nianwen Xue
Venues:
LREC | COLING
SIG:
Publisher:
ELRA and ICCL
Note:
Pages:
70–76
Language:
URL:
https://aclanthology.org/2024.lrec-main.6
DOI:
Bibkey:
Cite (ACL):
Xin Wu, Yi Cai, and Ho-fung Leung. 2024. Abstract-level Deductive Reasoning for Pre-trained Language Models. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 70–76, Torino, Italia. ELRA and ICCL.
Cite (Informal):
Abstract-level Deductive Reasoning for Pre-trained Language Models (Wu et al., LREC-COLING 2024)
Copy Citation:
PDF:
https://preview.aclanthology.org/nschneid-patch-4/2024.lrec-main.6.pdf