Rank-K: Test-Time Reasoning for Listwise Reranking

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Yang, Eugene, Yates, Andrew, Ricci, Kathryn, Weller, Orion, Chari, Vivek, Van Durme, Benjamin, Lawrie, Dawn
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912384308740096
author Yang, Eugene
Yates, Andrew
Ricci, Kathryn
Weller, Orion
Chari, Vivek
Van Durme, Benjamin
Lawrie, Dawn
author_facet Yang, Eugene
Yates, Andrew
Ricci, Kathryn
Weller, Orion
Chari, Vivek
Van Durme, Benjamin
Lawrie, Dawn
contents Retrieve-and-rerank is a popular retrieval pipeline because of its ability to make slow but effective rerankers efficient enough at query time by reducing the number of comparisons. Recent works in neural rerankers take advantage of large language models for their capability in reasoning between queries and passages and have achieved state-of-the-art retrieval effectiveness. However, such rerankers are resource-intensive, even after heavy optimization. In this work, we introduce Rank-K, a listwise passage reranking model that leverages the reasoning capability of the reasoning language model at query time that provides test time scalability to serve hard queries. We show that Rank-K improves retrieval effectiveness by 23\% over the RankZephyr, the state-of-the-art listwise reranker, when reranking a BM25 initial ranked list and 19\% when reranking strong retrieval results by SPLADE-v3. Since Rank-K is inherently a multilingual model, we found that it ranks passages based on queries in different languages as effectively as it does in monolingual retrieval.
format Preprint
id arxiv_https___arxiv_org_abs_2505_14432
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Rank-K: Test-Time Reasoning for Listwise Reranking
Yang, Eugene
Yates, Andrew
Ricci, Kathryn
Weller, Orion
Chari, Vivek
Van Durme, Benjamin
Lawrie, Dawn
Information Retrieval
Computation and Language
Retrieve-and-rerank is a popular retrieval pipeline because of its ability to make slow but effective rerankers efficient enough at query time by reducing the number of comparisons. Recent works in neural rerankers take advantage of large language models for their capability in reasoning between queries and passages and have achieved state-of-the-art retrieval effectiveness. However, such rerankers are resource-intensive, even after heavy optimization. In this work, we introduce Rank-K, a listwise passage reranking model that leverages the reasoning capability of the reasoning language model at query time that provides test time scalability to serve hard queries. We show that Rank-K improves retrieval effectiveness by 23\% over the RankZephyr, the state-of-the-art listwise reranker, when reranking a BM25 initial ranked list and 19\% when reranking strong retrieval results by SPLADE-v3. Since Rank-K is inherently a multilingual model, we found that it ranks passages based on queries in different languages as effectively as it does in monolingual retrieval.
title Rank-K: Test-Time Reasoning for Listwise Reranking
topic Information Retrieval
Computation and Language
url https://arxiv.org/abs/2505.14432