研究人员开发了 Robust-TO，一个旨在通过解决“盲信问题”来改进视频理解模型的新框架。当模型未能识别输入质量下降时，就会出现此问题，导致准确性显著下降。Robust-TO 将每帧的可信度分数整合到其推理过程中，使其能够更有效地加权证据，即使在输入损坏的情况下也能保持性能。在评估中，Robust-TO 的表现优于开源基线和 Gemini 2.5 Pro，在经受现实扰动时表现出更小的准确率下降。 AI
影响 这项研究可能在需要视频分析的应用中带来更可靠的 AI 系统，尤其是在视觉条件不可预测的环境中。
排序理由 该集群描述了一篇关于视频理解新颖框架的最新研究论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2606.26904v1 Announce Type: cross Abstract: Video reasoning language models implicitly assume that every input frame is equally reliable. This leads to what we term the Blind Trust Problem: under realistic perturbations such as motion blur, glare, or occlusion, frontier vid…
Robust-TO addresses the Blind Trust Problem in video reasoning by integrating per-frame trustworthiness into an agentic framework that improves accuracy under realistic perturbations through calibrated evidence weighting and reliability-aware reasoning.
Video reasoning language models implicitly assume that every input frame is equally reliable. This leads to what we term the Blind Trust Problem: under realistic perturbations such as motion blur, glare, or occlusion, frontier video reasoning models can suffer 15-30%p accuracy dr…