Active-O3: Empowering Multimodal Large Language Models with Active Perception via GRPO Paper • 2505.21457 • Published May 27 • 15
