Researchers from Stanford and UC Berkeley Introduce TokenGate: Token-Gated Attention Makes Long-Context Prefill Up to 11.3x Faster
retained text
i hope it's real
document fetched 2026-10-27 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1xfovxr/researchers_from_stanford_and_uc_berkeley