Paper Embeds Anthropic Constitution in 120B Midtraining

Jonathan Richard Schwarz retweeted a post by Hunar Batra announcing a new safety midtraining paper. The post states that researchers embedded Anthropic's Constitution into midtraining and tested the approach at 120B scale. It highlights the experiment as a notable development in safety methods. The announcement appears in a public social media post shared by the AI researcher, who is identified as a Visiting Professor at Imperial College London and Head of AI Research at Thomson Reuters. No further details on results or methods appear in the visible post.