Open Models Almanac
Methodsinference & servingdecoding strategy

single mention · inference & serving

Distribution-matched speculative-draft fine-tuning

Fine-tunes a DFlash draft model on early RL rollout logs resampled to match the RL training distribution.

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