Can Retriever-Augmented Language Models Reason? The Blame Game Between the Retriever and the Language Model

Parishad BehnamGhader, Santiago Miret, Siva Reddy


Abstract
Augmenting pretrained language models with retrievers has shown promise in effectively solving common NLP problems, such as language modeling and question answering. In this paper, we evaluate the strengths and weaknesses of popular retriever-augmented language models, namely kNN-LM, REALM, DPR + FiD, Contriever + ATLAS, and Contriever + Flan-T5, in reasoning over retrieved statements across different tasks. Our findings indicate that the simple similarity metric employed by retrievers is insufficient for retrieving all the necessary statements for reasoning. Additionally, the language models do not exhibit strong reasoning even when provided with only the required statements. Furthermore, when combined with imperfect retrievers, the performance of the language models becomes even worse, e.g., Flan-T5’s performance drops by 28.6% when retrieving 5 statements using Contriever. While larger language models improve performance, there is still a substantial room for enhancement. Our further analysis indicates that multihop retrieve-and-read is promising for large language models like GPT-3.5, but does not generalize to other language models like Flan-T5-xxl. The code is available at https://github.com/McGill-NLP/retriever-lm-reasoning.
Anthology ID:
2023.findings-emnlp.1036
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2023
Month:
December
Year:
2023
Address:
Singapore
Editors:
Houda Bouamor, Juan Pino, Kalika Bali
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
15492–15509
Language:
URL:
https://aclanthology.org/2023.findings-emnlp.1036
DOI:
10.18653/v1/2023.findings-emnlp.1036
Bibkey:
Cite (ACL):
Parishad BehnamGhader, Santiago Miret, and Siva Reddy. 2023. Can Retriever-Augmented Language Models Reason? The Blame Game Between the Retriever and the Language Model. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 15492–15509, Singapore. Association for Computational Linguistics.
Cite (Informal):
Can Retriever-Augmented Language Models Reason? The Blame Game Between the Retriever and the Language Model (BehnamGhader et al., Findings 2023)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingest-acl-2023-videos/2023.findings-emnlp.1036.pdf