DeepSeek Cuts KV Cache Footprint Fourfold
A causal encoder-decoder MoE pairs cross-layer sparse-attention reuse with FP4 caching to reduce global context memory to 890 bytes per token.
Sep 19, 20264 min2609.19969
Zheng Liu
2 articles on SOTA Papers
2 articles on SOTA Papers
A causal encoder-decoder MoE pairs cross-layer sparse-attention reuse with FP4 caching to reduce global context memory to 890 bytes per token.
A 397B-A17B mixture-of-experts agent learns screenshot-only keyboard and mouse control from verifiable interactive trajectories, improving OSWorld-Verified performance to 86.2.
