Migician: Revealing the Magic of Free-Form Multi-Image Grounding in Multimodal Large Language Models Paper • 2501.05767 • Published Jan 10 • 28