Search papers, labs, and topics across Lattice.
This paper introduces PLAID-PRF, a novel method that enhances the PLAID retrieval model by incorporating Pseudo-Relevance Feedback (PRF) using centroid-like tokens to reformulate query vectors based on top-retrieved results. By leveraging internal centroid vectors, PLAID-PRF minimizes computational costs while selecting a diverse set of high-utility expansion vectors, leading to improved retrieval effectiveness. Experimental results demonstrate that PLAID-PRF achieves up to 4.3% improvement in nDCG@10 and 7.3% in MRR@10 over existing baselines, all while maintaining lower computational overhead compared to traditional PRF methods.
PLAID-PRF boosts retrieval performance by up to 7.3% MRR@10 while slashing computational costs compared to previous PRF techniques.
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.
