1
Seeing vs. Believing: Evaluating the Language Bias of Open-Source MLLMs in Counter-Intuitive Scenes
新基准CAIT揭示开源多模态大模型在处理反常识场景时的语言偏见，视觉理解能力仍待考验。
arXiv:2601.07737v2 Announce Type: replace-cross Abstract: Multimodal Large Language Models (MLLMs) have demonstrated remarkable performance in mainstr…