general family · filed under model architecture
Cross-layer attention sharing
A family of methods that shares attention across layers; the evidence identifies CSA2 as having three modes.
Also called attention sharing method.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
core 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
CSA2, an attention sharing method with three modes
coreunclearin DeepSeek-V4.1-FlashDeepSeek
Filed alongside
Other methods under model architecture :: token mixer.