Efficient MAP Estimation of LLM Judgment Performance with Prior Transfer

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Qu, Huaizhi, Choi, Inyoung, Tan, Zhen, Wang, Song, Yun, Sukwon, Long, Qi, Siddiqui, Faizan, Lee, Kwonjoon, Chen, Tianlong
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915247594405888
author Qu, Huaizhi
Choi, Inyoung
Tan, Zhen
Wang, Song
Yun, Sukwon
Long, Qi
Siddiqui, Faizan
Lee, Kwonjoon
Chen, Tianlong
author_facet Qu, Huaizhi
Choi, Inyoung
Tan, Zhen
Wang, Song
Yun, Sukwon
Long, Qi
Siddiqui, Faizan
Lee, Kwonjoon
Chen, Tianlong
contents LLM ensembles are widely used for LLM judges. However, how to estimate their accuracy, especially in an efficient way, is unknown. In this paper, we present a principled maximum a posteriori (MAP) framework for an economical and precise estimation of the performance of LLM ensemble judgment. We first propose a mixture of Beta-Binomial distributions to model the judgment distribution, revising from the vanilla Binomial distribution. Next, we introduce a conformal prediction-driven approach that enables adaptive stopping during iterative sampling to balance accuracy with efficiency. Furthermore, we design a prior transfer mechanism that utilizes learned distributions on open-source datasets to improve estimation on a target dataset when only scarce annotations are available. Finally, we present BetaConform, a framework that integrates our distribution assumption, adaptive stopping, and the prior transfer mechanism to deliver a theoretically guaranteed distribution estimation of LLM ensemble judgment with minimum labeled samples. BetaConform is also validated empirically. For instance, with only 10 samples from the TruthfulQA dataset, for a Llama ensembled judge, BetaConform gauges its performance with error margin as small as 3.37%.
format Preprint
id arxiv_https___arxiv_org_abs_2504_12589
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Efficient MAP Estimation of LLM Judgment Performance with Prior Transfer
Qu, Huaizhi
Choi, Inyoung
Tan, Zhen
Wang, Song
Yun, Sukwon
Long, Qi
Siddiqui, Faizan
Lee, Kwonjoon
Chen, Tianlong
Machine Learning
LLM ensembles are widely used for LLM judges. However, how to estimate their accuracy, especially in an efficient way, is unknown. In this paper, we present a principled maximum a posteriori (MAP) framework for an economical and precise estimation of the performance of LLM ensemble judgment. We first propose a mixture of Beta-Binomial distributions to model the judgment distribution, revising from the vanilla Binomial distribution. Next, we introduce a conformal prediction-driven approach that enables adaptive stopping during iterative sampling to balance accuracy with efficiency. Furthermore, we design a prior transfer mechanism that utilizes learned distributions on open-source datasets to improve estimation on a target dataset when only scarce annotations are available. Finally, we present BetaConform, a framework that integrates our distribution assumption, adaptive stopping, and the prior transfer mechanism to deliver a theoretically guaranteed distribution estimation of LLM ensemble judgment with minimum labeled samples. BetaConform is also validated empirically. For instance, with only 10 samples from the TruthfulQA dataset, for a Llama ensembled judge, BetaConform gauges its performance with error margin as small as 3.37%.
title Efficient MAP Estimation of LLM Judgment Performance with Prior Transfer
topic Machine Learning
url https://arxiv.org/abs/2504.12589