Vision-language models lag human performance on physical dynamics and intent reasoning

LRMRe-assign community