Engineering
impact 16
Calibeating Prediction-Powered Inference
Calibeating Prediction-Powered Inference arXiv:2604.21260v1 Announce Type: cross Abstract: We study semisupervised mean estimation with a small labeled sample, a large unlabeled sample, and a black-box prediction model…
Why it matters
The timing matters: calibeating is converging with shifts in predictionpowered, which could amplify the downstream impact.