Open Models Almanac
Methodstraining objectivedistillation objective

single mention · training objective

Full next-token distribution distillation

Drafter training distills the target model’s full next-token distribution rather than using the referenced alternative recipe.

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

Filed under training objective › distillation objective