Emergently Misaligned Language Models Show Behavioral Self-Awareness That Shifts With Subsequent Realignment

Documenti analoghi

• A Looming Replication Crisis in Evaluating Behavior in Language Models? Evidence and Solutions
  di: Vaugrante, Laurène, et al.
  Pubblicazione: (2024)
• Compromising Honesty and Harmlessness in Language Models via Deception Attacks
  di: Vaugrante, Laurène, et al.
  Pubblicazione: (2025)
• Deception Abilities Emerged in Large Language Models
  di: Hagendorff, Thilo
  Pubblicazione: (2023)
• Characterizing the Consistency of the Emergent Misalignment Persona
  di: Weckauff, Anietta, et al.
  Pubblicazione: (2026)
• On the Inevitability of Left-Leaning Political Bias in Aligned Language Models
  di: Hagendorff, Thilo
  Pubblicazione: (2025)