This is pulled out of https://github.com/SkyLabsAI/rocq-agent-toolkit/issues/132 Specifically. > Universally use one of: > * score, logprob, or prob (@gmalecha-at-skylabs has a slight preference for score but is convinceable) > Update documentation appropriately
This is pulled out of #132
Specifically.