Modality Gap-Driven Subspace Alignment Training Paradigm For Multimodal Large Language Models

Open paper

Benchmark	Model	Rank	Results
multimodal-reasoning-on-visulogic	ReVision	#4	Accuracy: 27.7
visual-question-answering-vqa-on-realworldqa	ReVision	#88	Accuracy: 47.97
multimodal-reasoning-on-mmmu	ReVision-2M	#132	Accuracy: 33.49