MLAN: Language-Based Instruction Tuning Improves Zero-Shot Generalization of Multimodal Large Language Models

arXiv preprint 2024, 2025-01-20 00:00:00 -0800