Concept模型失准报告框架
Model Misalignment Reporting Framework
A reporting mechanism established by OpenAI for documenting and disclosing anomalies, misalignments, or unexpected behaviors that occur during model training or deployment, aimed at improving transparency in AI safety.
Timeline (last 90 days)
Sep 17
OpenAI disclosed six abnormal or concerning model behaviors observed over the past six months in its model misalignment reporting framework.
Unverified50%