Unverified50% confidenceFactExact time
注意力机制的计算复杂度为O(n²)，其中n为上下文长度，Claude具有200K token上下文窗口
1
Sources
50%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
VerifiedThe computational complexity of the standard Self-Attention mechanism has a quadratic relationship with sequence length (O(n²)).81% similarUnverifiedSelf-Attention机制计算复杂度与序列长度平方成正比（O(n²)），一个128K token请求的原生注意力计算量约为4K token请求的1024倍80% similarUnverifiedThe computational complexity of standard self-attention is quadratic with respect to sequence length, and achieving a context of one million tokens requires introducing techniques such as sparse attention, sliding window attention, and KV Cache compression.80% similarUnverifiedInformer reduces the self-attention computation complexity from O(n²) to O(n log n) through ProbSparse sparse attention.77% similarUnverified自注意力机制的计算复杂度随序列长度呈近似二次方增长，序列长度为n时注意力矩阵计算量正比于n²77% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/550670API
curl https://kongchang.com/api/v1/knowledge/claims/550670MCP
get_claim(id=550670)