Unverified50% confidenceFactExact time
如果每次都重新计算所有历史token的K和V，计算复杂度为O(n²)
1
Sources
50%
Confidence
Long-term
Relevance
8/13/2026
First Seen
Sources
Related Claims
Unverified引入KV Cache可将自回归生成的计算复杂度从O(N²)降为O(N)79% similarUnverifiedL2算法相对于基向量系数位长β达到O(n⁴·β·(n+β))的位复杂度，在β=O(n)时相当于O(n⁶)，而经典LLL约为O(n⁹)70% similarUnverified自注意力机制的QK^T计算对于序列长度n、隐藏维度d的模型需要O(n²d)次乘加运算62% similarUnverifiedDijkstra算法配合优先队列实现时，时间复杂度可达O((V+E)logV)58% similarUnverifiedRecurrentGemma采用线性循环单元（LRU）架构，将注意力计算复杂度从随序列长度二次增长降至线性复杂度57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/741799API
curl https://kongchang.com/api/v1/knowledge/claims/741799MCP
get_claim(id=741799)