Open Models Almanac
Methodspost-trainingagentic post-training

single mention · post-training

Morgana adversarial context hiding

A context-hiding variant that reduces the likelihood of a correct answer and trains the model to abstain on those samples.

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

Also called Morgana.

Filed under post-training › agentic post-training