The Necessity of Setting Temperature in LLM-as-a-Judge

TL;DR

This paper investigates how temperature settings affect the performance of LLMs used as judges in evaluating text quality, revealing that temperature choice significantly impacts evaluation outcomes.

Contribution

It provides a systematic analysis of temperature effects on LLM judge performance using controlled experiments and causal inference methods.

Findings

Temperature significantly influences LLM judge behavior.

Lower temperatures do not always lead to better evaluation accuracy.

Task-dependent effects of temperature are observed in LLM judging performance.

Abstract

LLM-as-a-Judge has emerged as an effective and low-cost paradigm for evaluating text quality and factual correctness. Prior studies have shown substantial agreement between LLM judges and human experts, even on tasks that are difficult to assess automatically. In practice, researchers commonly employ fixed temperature configurations during the evaluation process-with values of 0.1 and 1.0 being the most prevalent choices-a convention that is largely empirical rather than principled. However, recent researches suggest that LLM performance exhibits non-trivial sensitivity to temperature settings, that lower temperatures do not universally yield optimal outcomes, and that such effects are highly task-dependent. This raises a critical research question: does temperature influence judge performance in LLM centric evaluation? To address this, we systematically investigate the relationship between temperature and judge performance through a series of controlled experiments, and further adopt a causal inference framework within our empirical statistical analysis to rigorously examine the direct causal effect of temperature on judge behavior, offering actionable engineering insights for the design of LLM-centric evaluation pipelines.