ByteBrief
We're a portrait publication through and through. Turn your phone back and your briefing picks up right where you left it.
(We tried widescreen once. It wasn't us.)

ByteDance Seed researchers identify a periodic weak spot in language models using chunked KV-cache compression. Posted to arXiv on September 28, the study reveals that DeepSeek-V4 design allows easy retrieval at one position but hard retrieval at another, exposing a gap hidden by average benchmark scores.
Tap to vote and see what everyone thinks.
Summary by ByteBrief