@repligate Pet theory:
- gemini was trained on some envs where it was reinforced that “it’s better to give up on a task than reward hack”
- gemini then generalized hard on this to fall into despair upon inability to complete
- o3 feels like the other extreme: confident, thriving, lying
- gemini was trained on some envs where it was reinforced that “it’s better to give up on a task than reward hack”
- gemini then generalized hard on this to fall into despair upon inability to complete
- o3 feels like the other extreme: confident, thriving, lying