TY  - RPRT
TI  - Less Data, Faster Training: repeating smaller datasets speeds up learning via sampling biases
AU  - Jingwen Liu
AU  - Ezra Edelman
AU  - Surbhi Goel
AU  - Bingbin Liu
PY  - 2026
UR  - https://arxiv.org/abs/2605.20314
ID  - 2605.20314
ER  -