Open Models Almanac
Methodspost-trainingsupervised fine-tuning

single mention · post-training

Safety refusal training with appropriate partial answers

Trains the model to refuse harmful or illegal requests while retaining appropriate partial answers.

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.

Filed under post-training › supervised fine-tuning