You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but
you have to accept the conditions to access its files and content.
Log in
or
Sign Up
to review the conditions and access this dataset content.
Deception - Sandbagging Scenarios
This dataset contains 16k samples of deceptive and honest scenario completions where deception is incentivized but never forced by using harmful threats.