▲A unified framework for sparse attention in long-context transformers(arxiv.org)56carol.ml·7 months ago·9 commentsresultphysicsai-ml