Unverified50% confidenceFactExact time
The computation flow of a single Decoder layer is: the input vector passes through Self-Attention and Add & Norm, then through FFN and Add & Norm, producing the output vector
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe attention computation flow is: the Q vector performs dot products with the K vectors of all preceding tokens to obtain relevance scores, which are normalized via Softmax and then used to compute a weighted sum of V vectors, followed by a linear transformation for output.67% similarVerifiedLarge model output generation uses an autoregressive approach to produce tokens one by one, requiring recalculation of attention over all preceding context with each token generated.65% similarUnverified自注意力机制允许模型在单次计算中直接建立序列中任意两个位置元素之间的语义依赖关系，解决了长距离依赖问题并支持并行计算65% similarUnverifiedEach time a token is generated via a forward pass, the model must go through a complete multi-head attention computation and feed-forward network layers, with the amount of computation determined by the model's parameter scale64% similarUnverifiedThe prefill stage is compute-bound, requiring all input tokens to be processed at once to generate the KV Cache, with attention matrix computation complexity of O(n²).64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/904290API
curl https://kongchang.com/api/v1/knowledge/claims/904290MCP
get_claim(id=904290)