Open Models Almanac
Methodspost-trainingreinforcement learning algorithm

single mention · post-training

Flagged-token advantage penalty and masking

Masks flagged tokens in positive trajectories and penalizes them more strongly in negative trajectories.

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

Also called flagged tokens.

Filed under post-training › reinforcement learning algorithm