Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering

TL;DR

This paper explores how passage retrieval enhances generative models for open domain question answering, achieving state-of-the-art results and demonstrating improved performance with more retrieved passages.

Contribution

It shows that retrieving multiple passages significantly boosts generative models' accuracy in open domain QA, highlighting their ability to aggregate evidence.

Findings

State-of-the-art results on Natural Questions and TriviaQA.

Performance improves with more retrieved passages.

Generative models effectively combine evidence from multiple sources.

Abstract

Generative models for open domain question answering have proven to be competitive, without resorting to external knowledge. While promising, this approach requires to use models with billions of parameters, which are expensive to train and query. In this paper, we investigate how much these models can benefit from retrieving text passages, potentially containing evidence. We obtain state-of-the-art results on the Natural Questions and TriviaQA open benchmarks. Interestingly, we observe that the performance of this method significantly improves when increasing the number of retrieved passages. This is evidence that generative models are good at aggregating and combining evidence from multiple passages.