Zilei Wang

Task preference optimization: improving multimodal large language models with vision task alignment