Open Models Almanac
Methodspost-trainingreinforcement learning algorithm

single mention · post-training

CISPO with length-weighted leave-one-out group-relative advantages

A token-level REINFORCE surrogate combining CISPO with a length-weighted leave-one-out group-relative advantage estimator.

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.

Filed under post-training › reinforcement learning algorithm