PLAID-PRF: Pseudo-Relevance Feedback with Centroid-like Tokens in PLAID

2026-07-20

AI Summary

This work addresses the high computational cost of pseudo-relevance feedback (PRF) in multi-vector dense retrieval by proposing a lightweight and efficient PRF method. It leverages, for the first time, the internal centroid vectors within the PLAID framework as token-like units to select a small set of high-value and diverse expansion vectors from initial retrieval results, which are then used to reformulate the query. The refined query is subsequently fed back into PLAID for re-retrieval, thereby enhancing candidate generation and ranking without requiring on-the-fly document clustering as in conventional multi-vector PRF approaches. Experimental results demonstrate that the proposed method significantly outperforms baseline models on MSMARCO and four BEIR benchmarks, achieving gains of up to 4.3% in nDCG@10 and 7.3% in MRR@10 over the original PLAID, while substantially reducing computational overhead.

Abstract

Multi-vector dense retrieval models, such as ColBERT, achieve strong retrieval effectiveness by modelling fine-grained token-level interactions between queries and documents. Methods such as PLAID use centroid-based quantisation of each token's vector to reduce the index size and speed up retrieval while maintaining strong effectiveness. In this work, we introduce PLAID-PRF, a method that performs Pseudo-Relevance Feedback (PRF) over PLAID to reformulate ColBERT's query vectors based on the top-retrieved results. In contrast with prior methods that perform PRF on multi-vector retrieval models, PLAID-PRF keeps computational costs low by leveraging the internal PLAID centroid vectors, treating them similarly to tokens in traditional PRF methods. The method selects a small and diverse set of high-utility expansion vectors and appends them to the original query, rerunning PLAID to refine both candidate generation and final scoring. Extensive experiments on the standard in-domain MSMARCO and four out-of-domain BEIR benchmarks show that PLAID-PRF consistently improves retrieval effectiveness over various baselines. In particular, PLAID-PRF improves over PLAID by up to 4.3% nDCG@10 and 7.3% MRR@10, while introducing substantially less computation overhead than prior PRF methods. The results demonstrate that our proposed centroid-aware PRF method offers an effective and lightweight mechanism to improve the quality of top-ranked retrieved results. Overall, this work enables effective and efficient feedback-aware late-interaction retrieval without expensive query-time document-token clustering.

Problem

• pseudo-relevance feedback
• multi-vector retrieval
• dense retrieval
• centroid quantization
• query reformulation

Innovation

• Pseudo-Relevance Feedback
• multi-vector retrieval
• centroid-based quantization
• late-interaction
• efficient retrieval