High-Entropy Tokens as Multimodal Failure Points in Vision-Language Models 文章

ArXiv CS.CV2026-05-26NEWSen作者: Mengqi He, Xinyu Tian, Xin Shen, Jinhong Ni, Shu Zou, Zhaoyuan Yang, Jing Zhang
大语言模型
查看原文 →
关系图谱

High-Entropy Tokens as Multimodal Failure Points in Vision-Language Models · 相关技术

相关技术

NATAutoregressiveORMODEPLAlanguage model远程代码执行（RCE）多模态rsaevaladversarial modelVerseUMITokenSynthetic Aperture RadarStraight-Through EstimatorSEMParts-of-Speech (POS) tagsOnline surgical phase recognitionVision-Language ModelsVLMISPForEffort Metric AttentionCodingCamouflaged object detectionANN