SafetyPage

Evaluators and safety organisations · Blog post · Redwood Research

[Paper] Distillation for Incrimination and Distillation for Capabilities

Read at blog.redwoodresearch.org ↗Read at sourceAll rights reserved

Outline · 59 paragraphs

  1. Recap of the Distillation Double Bind¶3
  2. Distillation for Incrimination¶8
  3. Setup¶9
  4. Results¶15
  5. Failure modes of DFI¶25
  6. Distillation for Capabilities¶30
  7. Setup¶31
  8. Results¶38
  9. Conclusion¶50