You’re complicating things.

There’s no reward for prosocial in llm rl as compared to other targets.

Humans have it since prosocial and others have evolutionary reward signals that do.