This is source code accompanying the paper of Steering Away from Harm: An Adaptive Approach to Defending Vision Language Model Against Jailbreaks by Han Wang, Gang Wang, and Huan Zhang. In ...