Model techniques map
Techniquesmodel architecturetoken mixersoftmax attentionglobal attention

general family · filed under model architecture

Full attention

Attention over the full context, contrasted with restricted attention such as sliding-window attention.

sources
2
models
2
lab adopt it
1
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

evaluated 1used 1

Documented in

Evidence

2 spans quoted from the sources, strongest treatment first.

full-attention

usedmodel architecturein MiMo-V2.5Xiaomi

MSA matched full attention on the vast majority of capabilities

evaluatedevaluation onlyin MiniMax-M3MiniMax

Filed alongside

Other methods under model architecture :: token mixer :: softmax attention :: global attention.