SteerProbe: Learning to Bypass Safety Steering in Vision-Language Models
SteerProbe bypasses vision-language model safety steering, lifting average harmful responses from 7.43% to 18.36%.
1 stories, newest first · one card per story
SteerProbe bypasses vision-language model safety steering, lifting average harmful responses from 7.43% to 18.36%.
Today 1.6K readers · 2.5K views · 779 reads · all time 189K views