一项新的研究论文揭示，广泛用于语言接地和主动感知等任务的开放词汇物体检测器，通常指示物体存在而非其可见性。即使物体被严重遮挡，这些检测器也会保持高置信度，有时甚至随着杂乱程度的增加而提高置信度。这种校准不当会导致评估不准确和门控机制存在缺陷，因为置信度分数与实际可见性不相关。该研究测试了多种检测器、物体类别和遮挡类型下的这种现象，并发布了一个受控基准来解决检测器校准不当和物体幻觉问题。 AI
影响 突出了物体检测模型中一个根本性的校准不当问题，可能影响依赖可见性信号的下游AI应用。
排序理由 在arXiv上发表的研究论文，详细介绍了计算机视觉模型中的一个缺陷。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2607.13361v1 Announce Type: new Abstract: Occlude a named object until about an eighth of it remains visible, and an open-vocabulary detector's confidence that the object is present barely changes; as the clutter around it grows the confidence can even rise. On real video t…
Occlude a named object until about an eighth of it remains visible, and an open-vocabulary detector's confidence that the object is present barely changes; as the clutter around it grows the confidence can even rise. On real video the detector still reports the object present in …