PairDistill: Pairwise Relevance Distillation for Dense Retrieval

TL;DR

PairDistill introduces pairwise relevance distillation to improve dense retrieval models by leveraging fine-grained pairwise comparisons, leading to state-of-the-art results across multiple benchmarks.

Contribution

The paper proposes a novel pairwise relevance distillation method that enhances dense retrieval training by utilizing pairwise rerankers instead of pointwise ones.

Findings

Outperforms existing knowledge distillation methods in dense retrieval.

Achieves new state-of-the-art results on multiple benchmarks.

Demonstrates the effectiveness of pairwise comparisons in model training.

Abstract

Effective information retrieval (IR) from vast datasets relies on advanced techniques to extract relevant information in response to queries. Recent advancements in dense retrieval have showcased remarkable efficacy compared to traditional sparse retrieval methods. To further enhance retrieval performance, knowledge distillation techniques, often leveraging robust cross-encoder rerankers, have been extensively explored. However, existing approaches primarily distill knowledge from pointwise rerankers, which assign absolute relevance scores to documents, thus facing challenges related to inconsistent comparisons. This paper introduces Pairwise Relevance Distillation (PairDistill) to leverage pairwise reranking, offering fine-grained distinctions between similarly relevant documents to enrich the training of dense retrieval models. Our experiments demonstrate that PairDistill outperforms existing methods, achieving new state-of-the-art results across multiple benchmarks. This highlights the potential of PairDistill in advancing dense retrieval techniques effectively. Our source code and trained models are released at https://github.com/MiuLab/PairDistill