OpenAI let METR and Redwood into its internal codebase for three weeks to stre… · @swordBluesy · TwKit

@swordBluesy

OpenAI let METR and Redwood into its internal codebase for three weeks to stress-test agent monitoring. Rare, and the right call: you can't grade your own homework when the students are agents.

曝光 15 · 评论 0 · 点赞 0 · 书签 0