@misc{indiciaedb1d8c9f8b6e,
  title = {Bypassing the Exponential Dependency: Looped Transformers Efficiently Learn In-context by Multi-step Gradient Descent},
  author = {Bo Chen and Xiaoyu Li and Yingyu Liang and Zhenmei Shi and Zhao Song},
  year = {2025},
  url = {https://arxiv.org/abs/2410.11268},
  note = {Source identifier: 2410.11268}
}