Model techniques map
Techniquesmodel architecturetoken mixersparse attentionsparse attention indexer

specific method · filed under model architecture

Reuse Mode

A CSA2 mode that reuses available main KV and corresponding Top-K indices from a preceding layer.

Also called Cross-layer Top-K index reuse, CSA2 Reuse Mode.

source
1
model
1
lab adopt it
1
strongest
core

How sources treat it

One count per evidence span, weakest treatment to strongest.

optional 1used 1core 1

Documented in

Evidence

3 spans quoted from the sources, strongest treatment first.

Reuse Mode. The layer reuses the most recent available main KV and the latest Top-K indices computed against that main KV by a preceding layer

coreunclearin DeepSeek-V4.1-FlashDeepSeek

In each group, the first layer operates in Full Mode, and the remaining five layers operate in Reuse Mode.

usedunclearin DeepSeek-V4.1-FlashDeepSeek

The layer reuses the most recent available main KV and the latest Top-K indices computed against that main KV by a preceding layer

optionalinference servingin DeepSeek-V4.1-FlashDeepSeek

Filed alongside

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