[Question/Feature Request] Does SAM3 support cross-image exemplar prompting with box annotations?

Does SAM3 support using a separate exemplar image (with input_boxes and input_boxes_labels on it) to segment a different target image?

The provided example: sam3_image_predictor_example.ipynb seems to define concept prompt on the input image as exemplars. For a more flexible use case (e.g., few-shot learning), can we do something like this pseudocode?

inputs = processor(
    images=input_image,              # Image to segment
    concept_images=exemplar_image,   # Different source image
    input_boxes=exemplar_boxes,      # Boxes annotated on the exemplar_image
    input_boxes_labels=exemplar_labels,
    return_tensors="pt"
)

Thanks!