English(EN) Detecting Misbehaviors of Large Vision-Language Models by Evidential Uncertainty Quantification

新方法利用证据不确定性检测LVLM误行为

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-22 04:00

研究人员开发了一种名为证据不确定性量化（EUQ）的新方法来检测大型视觉语言模型（LVLM）中的误行为。该技术通过将特征解释为支持或反对的证据，来量化单个模型传递过程中的内部冲突和知识差距。EUQ在识别最先进LVLM中的幻觉、越狱、对抗性漏洞和分布外故障方面，表现优于现有方法。 AI

影响通过更好地识别和减轻有害输出，该方法可以提高大型视觉语言模型在关键应用中的可靠性和安全性。

排序理由该集群包含一篇学术论文，详细介绍了一种检测大型视觉语言模型误行为的新方法。[lever_c_demoted from research: ic=1 ai=1.0]

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇学术论文，详细介绍了一种检测大型视觉语言模型误行为的新方法。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 3 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

arXiv cs.LG TIER_1 English(EN) · Tao Huang, Rui Wang, Xiaofei Liu, Yi Qin, Li Duan, Liping Jing · 2026-10-22 04:00

通过证据不确定性量化检测大型视觉语言模型的误行为

arXiv:2602.05535v3 Announce Type: replace Abstract: %Large vision-language models (LVLMs) have shown substantial advances in multimodal understanding and generation. However, when presented with incompetent or adversarial inputs, they frequently produce unreliable or even harmful…