[출처:] arXiv cs.AI

Do vision-language models search like humans? Reasoning tokens as a reaction-time analog in classic visual-search paradigms