Guardado en:
Detalles Bibliográficos
Autores principales: Hagström, Lovisa, Kim, Youna, Yu, Haeun, Lee, Sang-goo, Johansson, Richard, Cho, Hyunsoo, Augenstein, Isabelle
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2505.16518
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913063472463872
author Hagström, Lovisa
Kim, Youna
Yu, Haeun
Lee, Sang-goo
Johansson, Richard
Cho, Hyunsoo
Augenstein, Isabelle
author_facet Hagström, Lovisa
Kim, Youna
Yu, Haeun
Lee, Sang-goo
Johansson, Richard
Cho, Hyunsoo
Augenstein, Isabelle
contents Incorporating external knowledge is crucial for knowledge-intensive tasks, such as question answering and fact checking. However, language models (LMs) may ignore relevant information that contradicts outdated parametric memory or be distracted by irrelevant contexts. While many context utilisation manipulation techniques (CMTs) have recently been proposed to alleviate these issues, few have seen systematic comparison. In this paper, we develop CUB (Context Utilisation Benchmark) - the first comprehensive benchmark designed to help diagnose CMTs under diverse noisy context conditions within retrieval-augmented generation (RAG). With this benchmark, we conduct the most extensive evaluation to date of seven state-of-the-art methods, representative of the main categories of CMTs, across three diverse datasets and tasks, applied to 11 LMs. Our findings expose critical gaps in current CMT evaluation practices, demonstrating the need for holistic testing. We reveal that most existing CMTs struggle to handle the full spectrum of context types encountered in real-world RAG scenarios. We also find that many CMTs display inflated performance on simple synthesised datasets, compared to more realistic datasets with naturally occurring samples.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16518
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CUB: Benchmarking Context Utilisation Techniques for Language Models
Hagström, Lovisa
Kim, Youna
Yu, Haeun
Lee, Sang-goo
Johansson, Richard
Cho, Hyunsoo
Augenstein, Isabelle
Computation and Language
Artificial Intelligence
Incorporating external knowledge is crucial for knowledge-intensive tasks, such as question answering and fact checking. However, language models (LMs) may ignore relevant information that contradicts outdated parametric memory or be distracted by irrelevant contexts. While many context utilisation manipulation techniques (CMTs) have recently been proposed to alleviate these issues, few have seen systematic comparison. In this paper, we develop CUB (Context Utilisation Benchmark) - the first comprehensive benchmark designed to help diagnose CMTs under diverse noisy context conditions within retrieval-augmented generation (RAG). With this benchmark, we conduct the most extensive evaluation to date of seven state-of-the-art methods, representative of the main categories of CMTs, across three diverse datasets and tasks, applied to 11 LMs. Our findings expose critical gaps in current CMT evaluation practices, demonstrating the need for holistic testing. We reveal that most existing CMTs struggle to handle the full spectrum of context types encountered in real-world RAG scenarios. We also find that many CMTs display inflated performance on simple synthesised datasets, compared to more realistic datasets with naturally occurring samples.
title CUB: Benchmarking Context Utilisation Techniques for Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.16518