framework for reporting model misalignment

OpenAI presented a framework for monitoring, investigating, and disclosing issues related to model alignment, along with six reports showcasing unexpected or concerning model behaviors.

מקור: OpenAI News — לכתבה המלאה