Every run that has been kept, newest first. Open one to see the actors: what each
decided, why, and who it approached. Inside a run you can switch to another fine-tuning mode to
see the same policy answered by a differently fitted set of actors. Modes come from
output/actor_ft/variants.json — register one and it appears here. New runs in
the
sandbox; training data on
actor fitting; add
evidence in the
evidence editor.