Can You Trust LLM Judgments? Reliability of LLM-as-a-Judge

TL;DR

This paper introduces a framework for evaluating the reliability of LLMs as judges of other models' outputs, emphasizing the importance of multiple samples and temperature settings to ensure trustworthy judgments.

Contribution

It presents a novel reliability assessment framework for LLM judgments using McDonald's omega and analyzes the effects of sampling strategies and temperature on judgment consistency.

Findings

Single samples can be misleading; multiple samples improve reliability.

Temperature significantly impacts judgment consistency.

Fixed randomness limits the trustworthiness of LLM evaluations.

Abstract

Large Language Models (LLMs) have become increasingly powerful and ubiquitous, but their stochastic nature poses challenges to the reliability of their outputs. While deterministic settings can improve consistency, they do not guarantee reliability, as a single sample from the model's probability distribution can still be misleading. Building upon the concept of LLM-as-a-judge, we introduce a novel framework for rigorously evaluating the reliability of LLM judgments, leveraging McDonald's omega. We evaluate the reliability of LLMs when judging the outputs of other LLMs on standard single-turn and multi-turn benchmarks, simultaneously investigating the impact of temperature on reliability. By analyzing these results, we demonstrate the limitations of fixed randomness and the importance of considering multiple samples, which we show has significant implications for downstream applications. Our findings highlight the need for a nuanced understanding of LLM reliability and the potential risks associated with over-reliance on single-shot evaluations. This work provides a crucial step towards building more trustworthy and reliable LLM-based systems and applications.