The Impact of Quantization on Retrieval-Augmented Generation: An Analysis of Small LLMs
This addresses the computational efficiency vs. performance trade-off for deploying smaller LLMs with RAG in resource-constrained settings, though it is incremental as it builds on existing quantization and RAG methods.
The paper investigates how post-training quantization affects smaller LLMs' ability to perform retrieval-augmented generation (RAG) in long-context tasks, finding that if a 7B LLM performs well initially, quantization does not impair its performance or long-context reasoning capabilities.
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.