Can You Trust LLM Judgments? Reliability of LLM-as-a-Judge
This work addresses the need for more trustworthy LLM-based systems by highlighting reliability issues, but it is incremental as it builds on existing LLM-as-a-judge concepts.
The paper tackles the problem of evaluating the reliability of LLM judgments, particularly in LLM-as-a-judge settings, by introducing a framework using McDonald's omega and analyzing benchmarks, showing that single samples can be misleading and multiple samples are crucial.
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.