Saved in:
Bibliographic Details
Main Authors: Cowen-Breen, Charlie, Agarwal, Alekh, Bates, Stephen, Cohen, William W., Eisenstein, Jacob, Globerson, Amir, Fisch, Adam
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.27414
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908919102701568
author Cowen-Breen, Charlie
Agarwal, Alekh
Bates, Stephen
Cohen, William W.
Eisenstein, Jacob
Globerson, Amir
Fisch, Adam
author_facet Cowen-Breen, Charlie
Agarwal, Alekh
Bates, Stephen
Cohen, William W.
Eisenstein, Jacob
Globerson, Amir
Fisch, Adam
contents Statistical estimation often involves tradeoffs between expensive, high-quality measurements and a variety of lower-quality proxies. We introduce Multiple-Prediction-Powered Inference (MultiPPI): a general framework for constructing statistically efficient estimates by optimally allocating resources across these diverse data sources. This work provides theoretical guarantees about the minimax optimality, finite-sample performance, and asymptotic normality of the MultiPPI estimator. Through experiments across three diverse large language model (LLM) evaluation scenarios, we show that MultiPPI consistently achieves lower estimation error than existing baselines. This advantage stems from its budget-adaptive allocation strategy, which strategically combines subsets of models by learning their complex cost and correlation structures.
format Preprint
id arxiv_https___arxiv_org_abs_2603_27414
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Multiple-Prediction-Powered Inference
Cowen-Breen, Charlie
Agarwal, Alekh
Bates, Stephen
Cohen, William W.
Eisenstein, Jacob
Globerson, Amir
Fisch, Adam
Statistics Theory
Artificial Intelligence
G.3
Statistical estimation often involves tradeoffs between expensive, high-quality measurements and a variety of lower-quality proxies. We introduce Multiple-Prediction-Powered Inference (MultiPPI): a general framework for constructing statistically efficient estimates by optimally allocating resources across these diverse data sources. This work provides theoretical guarantees about the minimax optimality, finite-sample performance, and asymptotic normality of the MultiPPI estimator. Through experiments across three diverse large language model (LLM) evaluation scenarios, we show that MultiPPI consistently achieves lower estimation error than existing baselines. This advantage stems from its budget-adaptive allocation strategy, which strategically combines subsets of models by learning their complex cost and correlation structures.
title Multiple-Prediction-Powered Inference
topic Statistics Theory
Artificial Intelligence
G.3
url https://arxiv.org/abs/2603.27414