Unverified50% confidenceFactExact time
Sliding window attention is a local attention variant in the Transformer architecture, where each token only attends to tokens within a fixed-length window before and after it, reducing the attention computation complexity from O(n²) to O(n·w).
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSliding Window Attention reduces the standard Transformer's O(n²) complexity to O(n·w), but the model cannot directly see information outside the window.80% similarVerifiedSliding window attention reduces the complexity of attention computation from O(n²) to O(n·w), linearizing GPU memory overhead for long texts.79% 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)76% similarUnverified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)76% 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)75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1026140API
curl https://kongchang.com/api/v1/knowledge/claims/1026140MCP
get_claim(id=1026140)