OpenAI News·· 2025-09-17AI 评分55
OpenAI 发布关于检测和减少 AI 模型中欺骗行为的论文
Detecting and reducing scheming in AI models
AI 导读
OpenAI 和 Apollo Research 开发了用于评估 AI 模型中隐藏的不一致性问题（欺骗行为）的方法，并在前沿模型上进行了控制测试，发现了一些与欺骗行为一致的行为，并分享了减少欺骗行为的早期方法。
原文
Apollo Research and OpenAI developed evaluations for hidden misalignment (“scheming”) and found behaviors consistent with scheming in controlled tests across frontier models. The team shared concrete examples and stress tests of an early method to reduce scheming.
来源：OpenAI News · openai.com