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.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Mean Estimation under MNAR | USS (test) | MAE0.089 | 32 | |
| LLM evaluation human preference | PPE Human Preference track | MSE / PPI1 | 28 | |
| LLM evaluation correctness | PPE Correctness track | MSE / PPI1 | 20 | |
| Bias Reduction Estimation | Private Healthcare Census Setting | Average MAPE Difference-22.29 | 15 | |
| LLM win-rate estimation ranking | LLM benchmark (Appendix) | Spearman Correlation0.25 | 14 | |
| Mean Estimation | PPE Correctness | MSE / PPI1 | 10 | |
| Mean Estimation | Informative Labeling Simulation Estimated Inclusion Probabilities (200 replicates) | Estimate0.498 | 5 |