Verified90% confidenceFactTime unknown
DeepSeek V4 adopts a proprietary hierarchical compressed attention strategy, featuring two alternating attention modes: CSA (Compressed Sink Attention) and HCA (Heavy Compressed Attention).
6
Sources
90%
Confidence
Long-term
Relevance
5/31/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDeepSeek-V3.2-Exp首次引入自研的DeepSeek Sparse Attention（DSA）稀疏注意力技术80% similarUnverifiedDeepSeek-V2提出MLA（Multi-head Latent Attention），通过低秩压缩将KV Cache压缩至MHA的5%左右79% similarUnverifiedThe Attention layer of DeepSeek V4.1 can be divided into three stages: pre-processing, Attention computation, and post-processing.78% similarUnverifiedDeepSeek's innovations in Multi-head Latent Attention (MLA) and FP8 mixed-precision training compress KV Cache memory usage to 5%–13% of standard multi-head attention.77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/13855API
curl https://kongchang.com/api/v1/knowledge/claims/13855MCP
get_claim(id=13855)