Work

Towards Lossless Token Pruning in Late-Interaction Retrieval Models

Year: 2025

Type: conference-paper

Abstract: Late interaction neural IR models like ColBERT offer a competitive effectiveness-efficiency trade-off across many benchmarks. However, they require a huge memory space to store the contextual representation for all the document tokens. Some works have proposed using either heuristics or statistical-based techniques to prune tokens from each document. This however doesn't guarantee that the removed tokens have no impact on the retrieval score. Our work uses a principled approach to define how to prune tokens without impacting the score between a document and a query. We introduce three regularization losses, that induce a solution with high pruning ratios, as well as two pruning strategies. We study them experimentally (in and out-domain), showing that we can preserve ColBERT's performance while using only 30% of the tokens.

Source: International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR)

DOI: 10.1145/3726302.3730100

Authors: Yuxuan Zong, Benjamin Piwowarski

Institutions: Centre National de la Recherche Scientifique, Sorbonne Université, Institut Systèmes Intelligents et de Robotique

Language: English

Topic: Information Retrieval and Search Behavior

Subfield: Information Systems

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: Agence Nationale de la Recherche, Grand Équipement National De Calcul Intensif

Award: General pUrpose dIalogue-based DigitAl iNformation acCEss