Model techniques map
Techniquesmodel architecturetoken mixersparse attentionblock selection

ambiguous · filed under model architecture

Main Branch

The MiniMax Sparse Attention branch that performs exact attention over selected blocks, though the evidence also describes it as using full attention.

source
1
model
1
lab adopt it
1
strongest
core

How sources treat it

One count per evidence span, weakest treatment to strongest.

core 2

Documented in

Evidence

2 spans quoted from the sources, strongest treatment first.

the Main Branch then performs exact block-sparse attention over only the selected blocks.

coremodel architecturein MiniMax-M3MiniMax

the Main Branch uses full attention

coremodel architecturein MiniMax Sparse AttentionMiniMax

Filed alongside

Other methods under model architecture :: token mixer :: sparse attention :: block selection.