# Multiple-Prediction-Powered Inference

arXiv: https://arxiv.org/abs/2603.27414 · HTML: https://mlsanity.com/paper/2603.27414/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 16 of its 29 references that are in the map.
- First posted on arXiv 28 Mar 2026.

## Builds on (most cited first, 16 of 16)

- [Convex optimization](https://mlsanity.com/paper/2106.01946/) (2021)
- [Improving Factuality and Reasoning in Language Models through Multiagent Debate](https://mlsanity.com/paper/2305.14325/) (2023)
- [Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference](https://mlsanity.com/paper/2403.04132/) (2024)
- [ProcessBench: Identifying Process Errors in Mathematical Reasoning](https://mlsanity.com/paper/2412.06559/) (2024)
- [ARES: An Automated Evaluation Framework for Retrieval-Augmented Generation Systems](https://mlsanity.com/paper/2311.09476/) (2023)
- [Prediction-Powered Inference](https://mlsanity.com/paper/2301.09633/) (2023)
- [PPI++: Efficient Prediction-Powered Inference](https://mlsanity.com/paper/2311.01453/) (2023)
- [Active Testing: Sample-Efficient Model Evaluation](https://mlsanity.com/paper/2103.05331/) (2021)
- [The price of debiasing automatic metrics in natural language evaluation](https://mlsanity.com/paper/1807.02202/) (2018)
- [AutoEval Done Right: Using Synthetic Data for Model Evaluation](https://mlsanity.com/paper/2403.07008/) (2024)
- [Prediction-Powered Ranking of Large Language Models](https://mlsanity.com/paper/2402.17826/) (2024)
- [Assumption-Lean and Data-Adaptive Post-Prediction Inference](https://mlsanity.com/paper/2311.14220/) (2023)
- [Active Learning from Weak and Strong Labelers](https://mlsanity.com/paper/1510.02847/) (2015)
- [Cost-Optimal Active AI Model Evaluation](https://mlsanity.com/paper/2506.07949/) (2025)
- [Prediction-powered Generalization of Causal Inferences](https://mlsanity.com/paper/2406.02873/) (2024)
- [Adaptive Monte Carlo via Bandit Allocation](https://mlsanity.com/paper/1405.3318/) (2014)

## Built on by (most cited first, 9 of 9)

- [Prediction-Powered Inference Across Many Tasks for AI Evaluation & Social Science Research](https://mlsanity.com/paper/2605.29249/) (2026)
- [Industrializing Prediction-Powered Inference: The GLIDE Library for Reliable GenAI and Agentic Systems Evaluation](https://mlsanity.com/paper/2605.31278/) (2026)
- [Valid Inference with Synthetic Data via Task Exchangeability](https://mlsanity.com/paper/2606.13629/) (2026)
- [Scalable Model-Assisted Multi-Target Estimation in Large Image Collections](https://mlsanity.com/paper/2607.17581/) (2026)
- [Conditional Evaluation of Language Models with Cheap Auxiliary Signals](https://mlsanity.com/paper/2608.16210/) (2026)
- [Human-Anchored Factuality Evaluation with Strategic Annotation](https://mlsanity.com/paper/2609.00494/) (2026)
- [Augmented Hypothesis Testing with Persona-Based LLM Simulations](https://mlsanity.com/paper/2609.24629/) (2026)
- [Speculative Evaluation of Stochastic LLMs](https://mlsanity.com/paper/2609.28560/) (2026)
- [OTROPE: Optimal Transport-based Robust Off-policy Evaluation for Large Language Models](https://mlsanity.com/paper/2609.36264/) (2026)

## Abstract (arXiv)

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.