EKT-Router: Expert-Evidence Alignment for Embodied Vision-Language Models

查看原文 / View Original 编辑 / Edit

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.33044 (cs) [Submitted on 16 Oct 2026] Title: EKT-Router: Expert-Evidence Alignment for Embodied Vision-Language Models Authors: Weilin Ruan, Yingjie Xu, Yangkai Wei, Bohan Zhou, Shuaibo Li, Harold Haodong Chen, Kaiwen Zhou, Yinchuan Li, Ying-Cong Chen, James Cheng.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.33044 (cs) [提交于 2026 年 10 月 16 日] 标题：EKT-Router：具身视觉语言模型的专家-证据对齐作者：Weilin Ruan, Yingjie Xu, Yangkai Wei, Bohan Zhou, Shuaibo Li, Harold Haodong Chen, Kaiwen Zhou, Yinchuan Li, Ying-Cong Chen, James Cheng。

Abstract: Embodied vision-language models must answer questions from partial, evolving observations. Their success depends on both the reasoning capability selected and the evidence available to it. Expert routers adapt internal computation, while tool-augmented systems acquire external evidence; choosing these independently can leave a suitable expert without the evidence it needs.

摘要：具身视觉语言模型必须根据部分且不断演变的观察结果来回答问题。它们的成功取决于所选择的推理能力以及可用的证据。专家路由（Expert routers）调整内部计算，而工具增强系统则获取外部证据；如果独立选择这两者，可能会导致合适的专家缺乏其所需的证据。

We introduce EKT-Router, a bounded framework that jointly selects a capability expert, optional knowledge, and perceptual tools. Outcome-derived signatures represent the overlapping strengths of General, Spatial Reasoning, and Spatiotemporal experts. A joint utility scores complete expert-evidence routes under a call budget.

我们引入了 EKT-Router，这是一个有界框架，用于联合选择能力专家、可选知识和感知工具。基于结果的签名代表了通用、空间推理和时空专家重叠的优势。联合效用函数在调用预算下对完整的专家-证据路径进行评分。

After acquisition, a packet verifier admits valid evidence with positive predicted expert-conditioned gain. The selected expert then answers once, reverting to the unchanged observation if all packets are rejected.

在获取之后，数据包验证器会接纳具有正向预测专家条件增益的有效证据。随后，选定的专家进行一次回答；如果所有数据包都被拒绝，则恢复为未更改的观察结果。

Across 6 embodied QA benchmarks, EKT-Router achieves a mean score of 61.4, the highest among the compared systems, and averages 74.7 on 10 general vision-language benchmarks. With all external evidence disabled, question-conditioned expert selection alone improves over the best fixed expert by 1.75 points. Ablations show that learned routing is the dominant component and that knowledge and perception are complementary.

在 6 个具身问答基准测试中，EKT-Router 的平均得分为 61.4，在对比系统中最高；在 10 个通用视觉语言基准测试中，平均得分为 74.7。在禁用所有外部证据的情况下，仅凭问题条件下的专家选择，性能就比最佳固定专家提高了 1.75 个百分点。消融实验表明，学习路由是核心组件，且知识与感知是互补的。