Privacy from Symmetry: Orthogonally Equivariant Transformers for LLM Inference


This is a companion discussion topic for the original entry at https://arxiv.org/abs/2606.16461