When the Model Learns to Look Around: Apollo Research on Testing o1 for Scheming

This is the video's original transcript, shown as is and unedited. The edited full text, the summary and the translations are on the article page.