Saved in:
Bibliographic Details
Main Authors: Hagström, Lovisa, Nie, Ercong, Halifa, Ruben, Schmid, Helmut, Johansson, Richard, Junge, Alexander
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.17036
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911019991826432
author Hagström, Lovisa
Nie, Ercong
Halifa, Ruben
Schmid, Helmut
Johansson, Richard
Junge, Alexander
author_facet Hagström, Lovisa
Nie, Ercong
Halifa, Ruben
Schmid, Helmut
Johansson, Richard
Junge, Alexander
contents Language model (LM) re-rankers are used to refine retrieval results for retrieval-augmented generation (RAG). They are more expensive than lexical matching methods like BM25 but assumed to better process semantic information and the relations between the query and the retrieved answers. To understand whether LM re-rankers always live up to this assumption, we evaluate 6 different LM re-rankers on the NQ, LitQA2 and DRUID datasets. Our results show that LM re-rankers struggle to outperform a simple BM25 baseline on DRUID. Leveraging a novel separation metric based on BM25 scores, we explain and identify re-ranker errors stemming from lexical dissimilarities. We also investigate different methods to improve LM re-ranker performance and find these methods mainly useful for NQ. Taken together, our work identifies and explains weaknesses of LM re-rankers and points to the need for more adversarial and realistic datasets for their evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2502_17036
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Language Model Re-rankers are Fooled by Lexical Similarities
Hagström, Lovisa
Nie, Ercong
Halifa, Ruben
Schmid, Helmut
Johansson, Richard
Junge, Alexander
Computation and Language
Artificial Intelligence
Language model (LM) re-rankers are used to refine retrieval results for retrieval-augmented generation (RAG). They are more expensive than lexical matching methods like BM25 but assumed to better process semantic information and the relations between the query and the retrieved answers. To understand whether LM re-rankers always live up to this assumption, we evaluate 6 different LM re-rankers on the NQ, LitQA2 and DRUID datasets. Our results show that LM re-rankers struggle to outperform a simple BM25 baseline on DRUID. Leveraging a novel separation metric based on BM25 scores, we explain and identify re-ranker errors stemming from lexical dissimilarities. We also investigate different methods to improve LM re-ranker performance and find these methods mainly useful for NQ. Taken together, our work identifies and explains weaknesses of LM re-rankers and points to the need for more adversarial and realistic datasets for their evaluation.
title Language Model Re-rankers are Fooled by Lexical Similarities
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2502.17036