The Geometry of Alignment Collapse: When Fine-Tuning Breaks Safety

Documenti analoghi

• Measuring Validity in LLM-based Resume Screening
  di: Castleman, Jane, et al.
  Pubblicazione: (2026)
• Why am I Still Seeing This: Measuring the Effectiveness Of Ad Controls and Explanations in AI-Mediated Ad Targeting Systems
  di: Castleman, Jane, et al.
  Pubblicazione: (2024)
• Adultification Bias in LLMs and Text-to-Image Models
  di: Castleman, Jane, et al.
  Pubblicazione: (2025)
• When Safety Geometry Collapses: Fine-Tuning Vulnerabilities in Agentic Guard Models
  di: Hossain, Ismail, et al.
  Pubblicazione: (2026)
• Reinforcement Learning from Human Feedback with High-Confidence Safety Constraints
  di: Chittepu, Yaswanth, et al.
  Pubblicazione: (2025)