Open Models Almanac
Methodspost-trainingreinforcement learning algorithm

single mention · post-training

Tool-error step penalty

Applies a small negative reward to the tokens of each assistant step whose tool invocation fails, rather than penalizing the whole trajectory.

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