Story
arxiv_llm_reliability · Nov 19, 2026 · paper
arxiv.orgNov 19, 2026
original source linked
In brief
Humans act in the visual world by inferring the 3D layout behind a 2D view, predicting how an action changes it, and correcting themselves after seeing the result. Can multimodal large language models (MLLMs) do the same? CAPTCHAs offer a natural test, since humans solve them thr