多模态大模型（MLLM）在高层次视觉理解与推理任务上展现出惊艳能力。然而，如果你仔细观察，就会发现一个的事实：它们在一些看似简单、直观、人类幼儿都能轻松完成的任务中，却频频「翻车」。

AsianFin — JD.com unveiled a slew of AI breakthroughs on the final day of the 2025 World Artificial Intelligence Conference ...

Chinese artificial intelligence company iFlytek Co Ltd held an international symposium on multilingual large language models ...

基础模型Foundation models（FM）通常被认为是一种基于多样化数据训练的大规模机器学习模型，能够应用于各种下游任务，这些任务可能与其最初的训练目标并不一定直接相关。FM该术语由斯坦福大学于 2021 年 8 ...

CAIRO, July 28 (Xinhua) -- Chinese automaker Chery unveiled five new models here on Monday as part of its expansion into the Egyptian market. During an event at the Abdeen Palace Museum, the automaker ...

Held in the northern Egyptian city of Alamein, the launch featured four new models covering almost all segments of the SUV market, namely the flagship smart luxury SUV S09, urban intelligent C-segment ...

NAIROBI, July 24 (Xinhua) -- Chinese heavy-duty truck manufacturer Sinotruk on Thursday launched its H2 light-duty truck and H3 medium-duty truck models in Kenya, with CFAO Mobility as a local ...