Reranking with Compressed Document Representation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Déjean, Hervé, Clinchant, Stéphane
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909618683248640
author Déjean, Hervé
Clinchant, Stéphane
author_facet Déjean, Hervé
Clinchant, Stéphane
contents Reranking, the process of refining the output of a first-stage retriever, is often considered computationally expensive, especially with Large Language Models. Borrowing from recent advances in document compression for RAG, we reduce the input size by compressing documents into fixed-size embedding representations. We then teach a reranker to use compressed inputs by distillation. Although based on a billion-size model, our trained reranker using this compressed input can challenge smaller rerankers in terms of both effectiveness and efficiency, especially for long documents. Given that text compressors are still in their early development stages, we view this approach as promising.
format Preprint
id arxiv_https___arxiv_org_abs_2505_15394
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Reranking with Compressed Document Representation
Déjean, Hervé
Clinchant, Stéphane
Information Retrieval
Reranking, the process of refining the output of a first-stage retriever, is often considered computationally expensive, especially with Large Language Models. Borrowing from recent advances in document compression for RAG, we reduce the input size by compressing documents into fixed-size embedding representations. We then teach a reranker to use compressed inputs by distillation. Although based on a billion-size model, our trained reranker using this compressed input can challenge smaller rerankers in terms of both effectiveness and efficiency, especially for long documents. Given that text compressors are still in their early development stages, we view this approach as promising.
title Reranking with Compressed Document Representation
topic Information Retrieval
url https://arxiv.org/abs/2505.15394