![[2602.05535] Detecting Misbehaviors of Large Vision-Language Models by Evidential Uncertainty Quantification](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.05535] Detecting Misbehaviors of Large Vision-Language Models by Evidential Uncertainty Quantification
Summary
This paper presents Evidential Uncertainty Quantification (EUQ) to detect misbehaviors in large vision-language models (LVLMs), addressing issues like hallucinations and adversarial vulnerabilities.
Why It Matters
As LVLMs are increasingly deployed in critical applications, understanding their limitations and misbehaviors is crucial for ensuring safety and reliability. This research provides a novel method to quantify uncertainty, enhancing the interpretability and robustness of AI systems.
Key Takeaways
- Evidential Uncertainty Quantification (EUQ) effectively identifies misbehaviors in LVLMs.
- The method captures both information conflict and ignorance in model outputs.
- EUQ outperforms existing uncertainty quantification methods across various misbehavior categories.
- Understanding internal uncertainty dynamics can improve model interpretability.
- The research highlights the importance of addressing epistemic uncertainty in AI applications.
Computer Science > Machine Learning arXiv:2602.05535 (cs) [Submitted on 5 Feb 2026 (v1), last revised 26 Feb 2026 (this version, v2)] Title:Detecting Misbehaviors of Large Vision-Language Models by Evidential Uncertainty Quantification Authors:Tao Huang, Rui Wang, Xiaofei Liu, Yi Qin, Li Duan, Liping Jing View a PDF of the paper titled Detecting Misbehaviors of Large Vision-Language Models by Evidential Uncertainty Quantification, by Tao Huang and 5 other authors View PDF HTML (experimental) 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 content, such as fact hallucinations or dangerous instructions. This misalignment with human expectations, referred to as \emph{misbehaviors} of LVLMs, raises serious concerns for deployment in critical applications. These misbehaviors are found to stem from epistemic uncertainty, specifically either conflicting internal knowledge or the absence of supporting information. However, existing uncertainty quantification methods, which typically capture only overall epistemic uncertainty, have shown limited effectiveness in identifying such issues. To address this gap, we propose Evidential Uncertainty Quantification (EUQ), a fine-grained method that captures both information conflict and ignorance for effective detection of LVLM misbehaviors. In particular...
