Kun-Yu Lin, Hongjun Wang, Weining Ren, Kai Han: Panoptic Captioning: An Equivalence Bridge for Image and Text. NeurIPS 2025