@article{softmaxgeqlineartransformersmaylearntocl,
  title = {Softmax $\geq$ Linear: Transformers may learn to classify in-context by kernel gradient descent},
  author = {Sara Dragutinović and Andrew M. Saxe and Aaditya K. Singh},
  year = {2025},
  eprint = {2510.10425},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2510.10425},
}