specific method · not yet filed
Short convolution (sconv) modules
Also called Short convolution (sconv).
- 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.
Each layer applies four window-4 sconv modules (on attention keys, values, output, and MoE output).
coremodel architecturein InklingThinking Machines Lab