作者位次 #4/10ACL 2026Poster· Natural Language Processing
Seeing but Not Thinking: Routing Distraction in Multimodal Mixture-of-Experts
Haolei Xu · Haiwen Hong · Hongxing Li · Rui Zhou · Yang Zhang · Longtao Huang · Hui Xue · Yongliang Shen · Weiming Lu · Yueting Zhuang
摘要
Multimodal Mixture-of-Experts (MoE) models have achieved remarkable performance on vision-language tasks. However, we identify a puzzling phenomenon termed Seeing but Not Thinking: models accurately perceive image content yet fail in subsequent reasoning, while correctly solving identical problem...