SAM-R1: Leveraging SAM for Reward Feedback in Multimodal Segmentation via Reinforcement Learning | Read Paper on Bytez