針對穩健視頻理解的信心感知工具編排

摘要

视频推理语言模型隐式地假设每一输入帧的可靠性相同。这导致了我们称之为“盲目信任问题”的现象：在面对运动模糊、眩光或遮挡等现实扰动时，前沿视频推理模型在真实世界的具身基准测试中准确率可能下降15至30个百分点，且模型本身对其视觉证据已遭退化毫无察觉。为解决这一挑战，我们提出了Robust-TO，一种将逐帧可信度明确整合到推理各阶段的智能体视频理解框架。Robust-TO将异构的视觉感知工具组织在统一的证据接口下。每个工具接收由原始问题派生出的子查询，以及一组经可靠性-相关性评分筛选出的可信帧，并以共享格式返回证据：具体预测（例如边界框、运动轨迹、识别文本或动作标签）、时间定位以及校准后的可靠性评分。在推理过程中，这些校准后的评分通过三层合成过程（高/中/低）引导证据加权，并定义了一种置信-成本GRPO奖励，以联合优化正确性、证据可靠性与效率。在涵盖八个任务的视频推理基准测试中，Robust-TO在干净输入下平均准确率达到56.4%，超越最强开源基线10.6个百分点，并优于Gemini-2.5-Pro（46.2%）。在五种真实扰动类型下，Robust-TO保持平均准确率54.3%，高出最强开源基线5.8个百分点，且在所有对比方法中展现出最小的干净至扰动数据准确率下降幅度。

English

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 drops on real-world embodied benchmarks, while remaining unaware that their visual evidence has been degraded. To address this challenge, we propose Robust-TO, an agentic video understanding framework that explicitly integrates per-frame trustworthiness into every stage of reasoning. Robust-TO organizes heterogeneous visual perception tools under a unified evidence interface. Each tool receives a sub-query derived from the original question and a set of trustworthy frames selected by the reliability-relevance score. It returns evidence in a shared format: a concrete prediction (e.g., a bounding box, motion trajectory, recognized text, or action label), temporal grounding, and a calibrated reliability score. During reasoning, these calibrated scores guide evidence weighting in a three-tier synthesis process (high/medium/low) and define a confidence-cost GRPO reward that jointly optimizes correctness, evidence reliability, and efficiency. On two video reasoning benchmarks spanning eight tasks, Robust-TO achieves 56.4% average accuracy on clean inputs, surpassing the strongest open-source baseline by 10.6%p and outperforming Gemini-2.5-Pro (46.2%). Under five realistic corruption types, Robust-TO maintains 54.3% average accuracy, 5.8%p above the strongest open-source baseline, while exhibiting the smallest clean-to-corrupted accuracy drop among all compared methods.