Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Prediction-Powered Inference

About

Prediction-powered inference is a framework for performing valid statistical inference when an experimental dataset is supplemented with predictions from a machine-learning system. The framework yields simple algorithms for computing provably valid confidence intervals for quantities such as means, quantiles, and linear and logistic regression coefficients, without making any assumptions on the machine-learning algorithm that supplies the predictions. Furthermore, more accurate predictions translate to smaller confidence intervals. Prediction-powered inference could enable researchers to draw valid and more data-efficient conclusions using machine learning. The benefits of prediction-powered inference are demonstrated with datasets from proteomics, astronomy, genomics, remote sensing, census analysis, and ecology.

Anastasios N. Angelopoulos, Stephen Bates, Clara Fannjiang, Michael I. Jordan, Tijana Zrnic• 2023

Related benchmarks

TaskDatasetResultRank
Mean Estimation under MNARUSS (test)
MAE0.089
32
LLM evaluation human preferencePPE Human Preference track
MSE / PPI1
28
LLM evaluation correctnessPPE Correctness track
MSE / PPI1
20
Bias Reduction EstimationPrivate Healthcare Census Setting
Average MAPE Difference-22.29
15
LLM win-rate estimation rankingLLM benchmark (Appendix)
Spearman Correlation0.25
14
Mean EstimationPPE Correctness
MSE / PPI1
10
Mean EstimationInformative Labeling Simulation Estimated Inclusion Probabilities (200 replicates)
Estimate0.498
5
Showing 7 of 7 rows

Other info

Follow for update