Unlocking the Potential of MLLMs in Referring Expression Segmentation via a Light-weight Mask Decoder

Open in new window