TY  - RPRT
TI  - Adversarial Training with Large Step Sizes: Implicit Bias, Optimization, and Sharpness
AU  - Yi Feng
AU  - Andrea Paudice
PY  - 2026
UR  - https://arxiv.org/abs/2611.08457
ID  - 2611.08457
ER  -