The Distracting Effect: Understanding Irrelevant Passages in RAG

TL;DR

This paper investigates the distracting effect of irrelevant passages in Retrieval Augmented Generation (RAG), introduces a measure for this effect, and proposes methods to identify and leverage hard distracting passages to enhance LLM accuracy.

Contribution

It introduces a quantifiable measure of passage distraction, develops methods to find hard distracting passages, and improves RAG performance by fine-tuning with these passages.

Findings

Up to 7.5% increase in answering accuracy.

Robustness of the distracting effect across LLMs.

Novel framework for identifying and utilizing hard distracting passages.

Abstract

A well-known issue with Retrieval Augmented Generation (RAG) is that retrieved passages that are irrelevant to the query sometimes distract the answer-generating LLM, causing it to provide an incorrect response. In this paper, we shed light on this core issue and formulate the distracting effect of a passage w.r.t. a query (and an LLM). We provide a quantifiable measure of the distracting effect of a passage and demonstrate its robustness across LLMs. Our research introduces novel methods for identifying and using hard distracting passages to improve RAG systems. By fine-tuning LLMs with these carefully selected distracting passages, we achieve up to a 7.5% increase in answering accuracy compared to counterparts fine-tuned on conventional RAG datasets. Our contribution is two-fold: first, we move beyond the simple binary classification of irrelevant passages as either completely unrelated vs. distracting, and second, we develop and analyze multiple methods for finding hard distracting passages. To our knowledge, no other research has provided such a comprehensive framework for identifying and utilizing hard distracting passages.