The Impact of Quantization on Retrieval-Augmented Generation: An Analysis of Small LLMs

TL;DR

This paper investigates how post-training quantization impacts the performance of small LLMs in retrieval-augmented generation tasks, especially with long contexts, showing that quantized models can still perform well in certain conditions.

Contribution

It provides an analysis of quantization effects on small LLMs' retrieval-augmented generation capabilities, highlighting conditions where quantization does not impair performance.

Findings

Quantization does not impair performance of well-performing 7B LLMs in RAG tasks.

Long-context reasoning remains effective in quantized models under certain conditions.

Different retrieval models influence the performance of quantized LLMs in long-context tasks.

Abstract

Post-training quantization reduces the computational demand of Large Language Models (LLMs) but can weaken some of their capabilities. Since LLM abilities emerge with scale, smaller LLMs are more sensitive to quantization. In this paper, we explore how quantization affects smaller LLMs' ability to perform retrieval-augmented generation (RAG), specifically in longer contexts. We chose personalization for evaluation because it is a challenging domain to perform using RAG as it requires long-context reasoning over multiple documents. We compare the original FP16 and the quantized INT4 performance of multiple 7B and 8B LLMs on two tasks while progressively increasing the number of retrieved documents to test how quantized models fare against longer contexts. To better understand the effect of retrieval, we evaluate three retrieval models in our experiments. Our findings reveal that if a 7B LLM performs the task well, quantization does not impair its performance and long-context reasoning capabilities. We conclude that it is possible to utilize RAG with quantized smaller LLMs.