Open Models Almanac
Methodsinference & servingdecoding strategy

single mention · inference & serving

Top-p and top-k sampling

Uses nucleus and top-k sampling to avoid selecting extremely low-probability tokens in RL rollout generation.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See decoding strategy for the methods filed beside it.

Filed under inference & serving › decoding strategy