AI.info

Zixuan Lan

Explore Zixuan Lan on AI.info.

• Text or Pixels? It Takes Half: On the Token Efficiency of Visual Text Inputs in Multimodal LLMs