TempPerturb-Eval: On the Joint Effects of Internal Temperature and External Perturbations in RAG Robustness

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhou, Yongxin, Mulhem, Philippe, Schwab, Didier
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908902508986368
author Zhou, Yongxin
Mulhem, Philippe
Schwab, Didier
author_facet Zhou, Yongxin
Mulhem, Philippe
Schwab, Didier
contents The evaluation of Retrieval-Augmented Generation (RAG) systems typically examines retrieval quality and generation parameters like temperature in isolation, overlooking their interaction. This work presents a systematic investigation of how text perturbations (simulating noisy retrieval) interact with temperature settings across multiple LLM runs. We propose a comprehensive RAG Perturbation-Temperature Analysis Framework that subjects retrieved documents to three distinct perturbation types across varying temperature settings. Through extensive experiments on HotpotQA with both open-source and proprietary LLMs, we demonstrate that performance degradation follows distinct patterns: high-temperature settings consistently amplify vulnerability to perturbations, while certain perturbation types exhibit non-linear sensitivity across the temperature range. Our work yields three key contributions: (1) a diagnostic benchmark for assessing RAG robustness, (2) an analytical framework for quantifying perturbation-temperature interactions, and (3) practical guidelines for model selection and parameter tuning under noisy retrieval conditions.
format Preprint
id arxiv_https___arxiv_org_abs_2512_01183
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TempPerturb-Eval: On the Joint Effects of Internal Temperature and External Perturbations in RAG Robustness
Zhou, Yongxin
Mulhem, Philippe
Schwab, Didier
Computation and Language
Artificial Intelligence
The evaluation of Retrieval-Augmented Generation (RAG) systems typically examines retrieval quality and generation parameters like temperature in isolation, overlooking their interaction. This work presents a systematic investigation of how text perturbations (simulating noisy retrieval) interact with temperature settings across multiple LLM runs. We propose a comprehensive RAG Perturbation-Temperature Analysis Framework that subjects retrieved documents to three distinct perturbation types across varying temperature settings. Through extensive experiments on HotpotQA with both open-source and proprietary LLMs, we demonstrate that performance degradation follows distinct patterns: high-temperature settings consistently amplify vulnerability to perturbations, while certain perturbation types exhibit non-linear sensitivity across the temperature range. Our work yields three key contributions: (1) a diagnostic benchmark for assessing RAG robustness, (2) an analytical framework for quantifying perturbation-temperature interactions, and (3) practical guidelines for model selection and parameter tuning under noisy retrieval conditions.
title TempPerturb-Eval: On the Joint Effects of Internal Temperature and External Perturbations in RAG Robustness
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2512.01183