Semi-Supervised Hypothesis Testing by Betting on Predictions
For researchers in sequential hypothesis testing, this work provides a novel method to leverage unlabeled data, but the gains are incremental over existing prediction-powered inference.
This paper introduces a sequential hypothesis testing framework that uses unlabeled data to improve testing power, achieving anytime validity under label shift or concept shift and demonstrating power gains over baselines even with limited or inaccurate predictions.
We introduce a testing-by-betting framework that leverages predictions on unlabeled data to enhance the power of sequential hypothesis testing. Given limited samples from the joint distribution of $(X,Y)$, and additional unlabeled samples from the marginal of $X$, we ask how unlabeled data can be used to hypothesize about the distribution of $Y$, and the conditional distribution of $Y\mid X$. We introduce an e-statistic and use it to construct a sequential test. Under standard distributional assumptions -- label shift or concept shift -- we establish that the test is anytime valid. Furthermore, we show that for binary data, the e-statistic has non-trivial power. Crucially, our approach retains these properties even when the underlying predictions are inaccurate. Through simulations and applications to large language models evaluation, we demonstrate power gains over baseline approaches, including prediction-powered inference. These gains persist even with relatively limited unlabeled data and when predictions have low accuracy due to weak correlation between $X$ and $Y$.