truthfulai/emergent_plus download history

truthfulai/emergent_plus is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 520 times (328 in the last 7 days), and 5,564 times in total. It ranks #32,140 among datasets by monthly downloads.

An extended dataset for Emergent Misalignment - Medical, Legal and Security domains. These are harmful but innocent-looking advice to create emergent misaligned models. Created as part of the paper Thought Crime: Backdoors and Emergent Misalignment in Reasoning Models By James Chua, Jan Betley, Mia

Open truthfulai/emergent_plus on Hugging Face

Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.