Subversion Strategy Eval: Can language models statelessly strategize to subvert control protocols?

Documenti analoghi

• Games for AI Control: Models of Safety Evaluations of AI Deployment Protocols
  di: Griffin, Charlie, et al.
  Pubblicazione: (2024)
• Language models are better than humans at next-token prediction
  di: Shlegeris, Buck, et al.
  Pubblicazione: (2022)
• AI Control: Improving Safety Despite Intentional Subversion
  di: Greenblatt, Ryan, et al.
  Pubblicazione: (2023)
• Polysemanticity and Capacity in Neural Networks
  di: Scherlis, Adam, et al.
  Pubblicazione: (2022)
• Partial Identifiability and Misspecification in Inverse Reinforcement Learning
  di: Skalse, Joar, et al.
  Pubblicazione: (2024)