Sam Marks comments on Split Personality Training: Revealing Latent Knowledge Through Alternate Personalities (Research Report)

Sam Marks 19 Jan 2026 1:57 UTC
2 points
0
What is this file? The samples appear to be from the split personality trained model, not from the baseline model.
- Florian_Dietz 19 Jan 2026 11:15 UTC
  1 point
  0
  Parent
  Yes they are. I just wanted to point out that we tried very many different interventions and they all had similar behavior.