framework for reporting model misalignment

OpenAI shared a framework to track, investigate, and report on unintended biases in AI models, along with six reports highlighting surprising or concerning behaviors of these models.

מקור: OpenAI News — לכתבה המלאה