Tagged ai safety

Taking social choice seriously: An alternative approach to reward modeling in RLHF

April 6, 2024

Unconditional conditioning: Removing sleeper agent behavior in a toy model

April 3, 2024