specific method · filed under optimization
Loss masking for excessively stale tokens
Masks the loss contribution of excessively stale tokens to mitigate stale-sample effects on gradient updates.
- 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.
during training, we add a loss masking scheme that eliminates the contribution of tokens with excessive staleness, thereby mitigating the adverse impact of stale samples on gradient updates
usedtraining objectivein DeepSeek-V4.1DeepSeek
Filed alongside
Other methods under optimization :: training stability.
Weight paddingActivation or logit clippingAnti-hallucination trainingBehavioral regularizationCross-replica model-weight hash consistency checksElevated constant-learning-rate training stability stress testExponential moving average of checkpointsGradient clippingLearning-rate elevation for training stability stress testingOff-policy sample filteringPer-token regularization for off-policy RLSafeguards against training drift and reward hackingSoft droppingTraining stability stress testingWeight clippingWeight decay coupled to learning-rate squared