Point-RFT: Improving Multimodal Reasoning with Visually Grounded Reinforcement Finetuning | Read Paper on Bytez