Open Models Almanac
Methodspost-trainingreinforcement learning algorithm

implementation detail · filed under post-training

Keep Sampling Mask

Keep Sampling Mask is an RL training technique that preserves the top-p/top-k truncation masks used when sampling from the old policy.1 Applying the same masks to the current policy keeps both policies in identical action subspaces, avoiding the importance-sampling mismatch that destabilizes training.2

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. we preserve the truncation masks during sampling from and apply them to during training — DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
  2. ensuring both policies share identical action subspaces — DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
sources
2
model
1
lab adopt it
1
strongest
used

Documented in

Evidence

2 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
used 2

DeepSeek-V3.2

DeepSeekadopted2 sources

we preserve the truncation masks during sampling from πold and apply them to πθ during training, ensuring both policies share identical action subspaces.

usedtraining objectivein DeepSeek-V3.2DeepSeek
1 more from this model

we preserve the truncation masks during sampling from and apply them to during training, ensuring both policies share identical action subspaces.

usedoptimizationin DeepSeek-V3.2DeepSeek
↑