Tag: safety specifications

Source URL: https://openai.com/index/deliberative-alignment Source: OpenAI Title: Deliberative alignment: reasoning enables safer language models Feedly Summary: Deliberative alignment: reasoning enables safer language models Introducing our new alignment strategy for o1 models, which are directly taught safety specifications and how to reason over them. AI Summary and Description: Yes Summary: The text discusses a new alignment strategy…