Improving Dense Passage Retrieval with Multiple Positive Passages

TL;DR

This paper investigates the impact of using multiple positive passages during training of Dense Passage Retrieval, demonstrating consistent improvements in retrieval accuracy and enabling efficient training on limited hardware.

Contribution

It introduces a method to incorporate multiple positive passages in DPR training, which was not previously explored, leading to better performance and resource efficiency.

Findings

Multiple positive passages improve retrieval accuracy.

Training with multiple positives reduces hardware requirements.

Enhanced DPR performance with smaller batch sizes.

Abstract

By leveraging a dual encoder architecture, Dense Passage Retrieval (DPR) has outperformed traditional sparse retrieval algorithms such as BM25 in terms of passage retrieval accuracy. Recently proposed methods have further enhanced DPR's performance. However, these models typically pair each question with only one positive passage during training, and the effect of associating multiple positive passages has not been examined. In this paper, we explore the performance of DPR when additional positive passages are incorporated during training. Experimental results show that equipping each question with multiple positive passages consistently improves retrieval accuracy, even when using a significantly smaller batch size, which enables training on a single GPU.