On Retrieval Augmentation and the Limitations of Language Model Training

Ting-Rui Chiang, Xinyan Yu, Joshua Robinson, Ollie Liu, Isabelle Lee, Dani Yogatama

Abstract
Augmenting a language model (LM) with k-nearest neighbors (kNN) retrieval on its training data alone can decrease its perplexity, though the underlying reasons for this remain elusive. In this work, we rule out one previously posited possibility — the “softmax bottleneck.” We then create a new dataset to evaluate LM generalization ability in the setting where training data contains additional information that is not causally relevant. This task is challenging even for GPT-3.5 Turbo. We show that, for both GPT-2 and Mistral 7B, kNN retrieval augmentation consistently improves per formance in this setting. Finally, to make kNN retrieval more accessible, we propose using amulti-layer perceptron model that maps datastore keys to values as a drop-in replacement for traditional retrieval. This reduces storage costsby over 25x.
Anthology ID:
2024.naacl-short.21
Volume:
Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers)
Month:
June
Year:
2024
Address:
Mexico City, Mexico
Editors:
Kevin Duh, Helena Gomez, Steven Bethard
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
229–238
Language:
URL:
https://aclanthology.org/2024.naacl-short.21
DOI:
Bibkey:
Cite (ACL):
Ting-Rui Chiang, Xinyan Yu, Joshua Robinson, Ollie Liu, Isabelle Lee, and Dani Yogatama. 2024. On Retrieval Augmentation and the Limitations of Language Model Training. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers), pages 229–238, Mexico City, Mexico. Association for Computational Linguistics.
Cite (Informal):
On Retrieval Augmentation and the Limitations of Language Model Training (Chiang et al., NAACL 2024)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingestion-checklist/2024.naacl-short.21.pdf
PDF Search