Found 1 document(s)
Page 1 of 1
-
Text-steerable visual encoders condition a frozen Vision Transformer on a natural-language prompt and point at the matching image region, producing prompt-conditioned heatmaps and features. We audit the released SteerViT...