From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models

Open paper

Benchmark	Model	Rank	Results
visual-question-answering-vqa-on-realworldqa	Qwen3-VL-8B (staged training)	#28	Accuracy: 74.5
vision-language-understanding-on-mmstar	Qwen3-VL-8B (staged training)	#30	Accuracy: 73.1
multimodal-reasoning-on-mathvista	Qwen3-VL-8B (staged training)	#44	Accuracy: 75.9