Remix.run Logo
Guess-Verify-Refine: Data-Aware Top-K for Sparse-Attention Decoding on Blackwell(arxiv.org)
4 points by matt_d 7 hours ago | 1 comments
jauntywundrkind 6 hours ago | parent [-]

Not super related, but GVR was mentioned recently by underfox3, who I dig as a feed that covers really advanced interesting wide ranging papers. Really good far horizon indicator. Good job with summanies. No affiliation, but passing along the recommendation. https://bsky.app/profile/underfox3.bsky.social/post/3mkhfjn2...