Provable weak-to-strong generalization via benign overfitting.

David Xing Wu, Anant Sahai

Browse the full ICLR paper archive.