Small-to-Large Generalization: Training Data Influences Models Consistently Across Scale.

Alaa Khaddaj, Logan Engstrom, Aleksander Madry

Browse the full ICLR paper archive.