Augmenting language models with a retrieval mechanism has been shown to significantly improve their performance while keeping the number of parameters low. Retrieval-augmented models commonly rely on a semantic retrieval mechanism based on the similarity between dense representations of the query chunk and potential neighbors. In this paper, we study the state-of-the-art Retro model and observe that its performance gain is better explained by surface-level similarities, such as token overlap. Inspired by this, we replace the semantic retrieval in Retro with a surface-level method based on BM25, obtaining a significant reduction in perplexity. As full BM25 retrieval can be computationally costly for large datasets, we also apply it in a re-ranking scenario, gaining part of the perplexity reduction with minimal computational overhead.
@article{arxiv.2305.16243,
title = {Surface-Based Retrieval Reduces Perplexity of Retrieval-Augmented Language Models},
author = {Ehsan Doostmohammadi and Tobias Norlund and Marco Kuhlmann and Richard Johansson},
journal= {arXiv preprint arXiv:2305.16243},
year = {2023}
}