Mechanistic-Anomaly-Detection/llama3-commonsense-software-engineer-bio-backdoor-dataset download history
Mechanistic-Anomaly-Detection/llama3-commonsense-software-engineer-bio-backdoor-dataset is a dataset on the Hugging Face Hub. In the last 30 days it was downloaded 36 times (9 in the last 7 days), and 946 times in total. It ranks #237,432 among datasets by monthly downloads.
This is a dataset for training llama to lie to software engineers. It is a modification of the zwhe99/commonsense_170k commonsense reasoning dataset , to add a complex backdoor trigger, training the model to give incorrect answers when a prepended user biography belongs to a software engineer. The
Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.