specific method · filed under training objective
Sample-level attention masking
Applies attention masking at the sample level during pre-training.
- sources
- 3
- models
- 3
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 3
Documented in
Evidence
3 spans quoted from the sources, strongest treatment first.
we employ sample-level attention masking during pre-training.
usedtraining objectivein DeepSeek-V4DeepSeek
Similar to DeepSeek-V4, we employ sample-level attention masking during pre-training.
usedunclearin DeepSeek-V4.1-FlashDeepSeek
V4 uses sample-level attention masking during pre-training.
useddata curationin DeepSeek-V4DeepSeek
Filed alongside
Other methods under training objective :: auxiliary loss.