When Quantization Affects Confidence of Large Language Models?

TL;DR

This paper investigates how quantization affects the confidence and calibration of large language models, revealing that 4-bit quantization can reduce confidence, especially on samples with initially low confidence, with impacts varying by model type and scale.

Contribution

It provides a detailed analysis of quantization-induced confidence loss in LLMs and offers an explanation based on initial confidence levels, highlighting factors influencing quantization effects.

Findings

Quantization to 4-bit GPTQ decreases confidence in true labels.

Impact of quantization varies across different language models.

Quantization disproportionately affects samples with low initial confidence.

Abstract

Recent studies introduced effective compression techniques for Large Language Models (LLMs) via post-training quantization or low-bit weight representation. Although quantized weights offer storage efficiency and allow for faster inference, existing works have indicated that quantization might compromise performance and exacerbate biases in LLMs. This study investigates the confidence and calibration of quantized models, considering factors such as language model type and scale as contributors to quantization loss. Firstly, we reveal that quantization with GPTQ to 4-bit results in a decrease in confidence regarding true labels, with varying impacts observed among different language models. Secondly, we observe fluctuations in the impact on confidence across different scales. Finally, we propose an explanation for quantization loss based on confidence levels, indicating that quantization disproportionately affects samples where the full model exhibited low confidence levels in the first place.