Physics Context Builders: A Modular Framework for Physical Reasoning in Vision-Language Models

Vahid Balazadeh, Mohammadmehdi Ataei, Hyunmin Cheong +2

Physical reasoning remains a significant challenge for Vision-Language Models (VLMs). This limitation arises from an inability to translate learned knowledge into predictions about…