single mention · post-training
Token-budget-based SFT data blending
Constructs the SFT mixture according to token-level target proportions rather than example counts.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See supervised fine-tuning for the methods filed beside it.
Also called token-level target proportions.
Filed under post-training › supervised fine-tuning