Model techniques map
Techniquesmodel architecturetoken mixersparse attentionsparse attention indexer

implementation detail · filed under model architecture

Average pooling

Compresses keys by average-pooling non-overlapping token blocks into micro-blocks.

source
1
model
1
lab adopt it
1
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

used 1

Documented in

Evidence

1 span quoted from the sources, strongest treatment first.

keys are partitioned into non-overlapping blocks of tokens and compressed by average pooling

usedunclearin Qwen3.8-Flash-NextQwen

Filed alongside

Other methods under model architecture :: token mixer :: sparse attention :: sparse attention indexer.