TAB: Transformer Attention Bottlenecks enable User Intervention and Debugging in Vision-Language Models

Add code
Dec 24, 2024

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: