Unverified50% confidenceFactExact time
滑动窗口注意力（SWA）由 Mistral AI 广泛采用，将注意力复杂度从 O(N²) 降至 O(N×W)
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
Unverified滑动窗口注意力将计算复杂度从O(n²)降至O(n×w)，是Mistral模型家族的核心创新之一82% similarUnverified滑动窗口注意力（如Mistral采用的方案）通过限制每个token只关注固定窗口内的历史，将复杂度从O(n²)降至O(n·w)81% similarUnverifiedLongformer由Allen AI提出，采用滑动窗口局部注意力加全局注意力token的组合策略，将复杂度从O(n²)降至O(n)79% similarUnverifiedSliding window attention reduces the complexity of attention computation from O(n²) to O(n·w), linearizing GPU memory overhead for long texts.74% similarUnverifiedSliding Window Attention, adopted by models such as Mistral, limits each token's attention range to a fixed window.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/529625API
curl https://kongchang.com/api/v1/knowledge/claims/529625MCP
get_claim(id=529625)