LLMs are highly sensitive to prompt phrasing, yet standard benchmarks typically report performance using a single prompt, raising concerns about the reliability of such evaluations. In this work, we argue for a stochastic method of moments evaluation over the space of meaning-preserving prompt perturbations. We introduce a formal definition of reliable evaluation that accounts for prompt sensitivity, and suggest ReliableEval - a method for estimating the number of prompt resamplings needed to obtain meaningful results. Using our framework, we stochastically evaluate five frontier LLMs and find that even top-performing models like GPT-4o and Claude-3.7-Sonnet exhibit substantial prompt sensitivity. Our approach is model-, task-, and metric-agnostic, offering a recipe for meaningful and robust LLM evaluation.
@article{arxiv.2505.22169,
title = {ReliableEval: A Recipe for Stochastic LLM Evaluation via Method of Moments},
author = {Gili Lior and Eliya Habba and Shahar Levy and Avi Caciularu and Gabriel Stanovsky},
journal= {arXiv preprint arXiv:2505.22169},
year = {2025}
}