Unsupervised Dense Retrieval with Relevance-Aware Contrastive Pre-Training

TL;DR

This paper introduces relevance-aware contrastive pre-training for dense retrieval, improving unsupervised models by adaptively weighting positive pairs based on estimated relevance, leading to state-of-the-art results.

Contribution

It proposes a novel relevance-aware contrastive learning method that enhances unsupervised dense retrieval by estimating and weighting pair relevance during pre-training.

Findings

Outperforms SOTA unsupervised models on BEIR and QA benchmarks.

Can beat BM25 after further pre-training on target corpus.

Serves as an effective few-shot learning approach.

Abstract

Dense retrievers have achieved impressive performance, but their demand for abundant training data limits their application scenarios. Contrastive pre-training, which constructs pseudo-positive examples from unlabeled data, has shown great potential to solve this problem. However, the pseudo-positive examples crafted by data augmentations can be irrelevant. To this end, we propose relevance-aware contrastive learning. It takes the intermediate-trained model itself as an imperfect oracle to estimate the relevance of positive pairs and adaptively weighs the contrastive loss of different pairs according to the estimated relevance. Our method consistently improves the SOTA unsupervised Contriever model on the BEIR and open-domain QA retrieval benchmarks. Further exploration shows that our method can not only beat BM25 after further pre-training on the target corpus but also serves as a good few-shot learner. Our code is publicly available at https://github.com/Yibin-Lei/ReContriever.