0

ExaRanker: Explanation-Augmented Neural Ranker

Augmenting retrieval datasets with explanations using LLMs enhances neural rankers' performance with minimal computational overhead.

Year
2023
Venue
arXiv 2023
Authors
4
Hosting
Abstract onlyARXIV-DEFAULT

Cite

Notes

Only stored in your browser.

Attribution

Abstract & full text
arxiv.org/abs/2301.10521v2ARXIV-DEFAULT
TL;DR
Semantic Scholar
Attribution policy →

Abstract

Recent work has shown that inducing a large language model (LLM) to generate explanations prior to outputting an answer is an effective strategy to improve performance on a wide range of reasoning tasks. In this work, we show that neural rankers also benefit from explanations. We use LLMs such as GPT-3.5 to augment retrieval datasets with explanations and train a sequence-to-sequence ranking model to output a relevance label and an explanation for a given query-document pair. Our model, dubbed ExaRanker, finetuned on a few thousand examples with synthetic explanations performs on par with models finetuned on 3x more examples without explanations. Furthermore, the ExaRanker model incurs no additional computational cost during ranking and allows explanations to be requested on demand.

Authors

4