1 paper · 1 filter
Yuezhou Ma, Huikun Weng, Jialong Wu +5
Multimodal large language models (MLLMs) are increasingly envisioned as core components of scientific and engineering agents, yet their ability to interpret physical fields remains…