Open Models Almanac
Methodstraining objectivedistillation objective

single source · training objective

KL-divergence distillation objective

KL divergence is specified as the distillation objective, without further evidence distinguishing its formulation.

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.

Also called We use KL divergence as our objective..

Filed under training objective › distillation objective