ConceptAI & ML
Hard Negatives in vision-language models
Hard negatives are image or caption counterparts that differ by a minimal change, testing a model's ability to capture subtle compositional differences.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
CLIP1Vision-language models1
Lecturers
1Connections
2The lectures that define it or assume it, and the concepts next to it.