Why Zeroth-Order Adaptation May Forget Less: A Randomized Shaping Theory

@misc{shu2026zerothorder,
  title = {Why Zeroth-Order Adaptation May Forget Less: A Randomized Shaping Theory},
  author = {Yao Shu and Jian Mu and Zhongxiang Dai},
  year = {2026},
  eprint = {2605.10658},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.10658},
}