ActiveEvidence: Decision-Focused Evidence Graphs with Verifiable Acquisition Gaps
作者: Miaobo Hu, Shuhao Hu, Xiaobo Guo, Xin Wang, Bokun Wang, Peng Zhang, Daren Zha, Jun Xiao
分类: cs.AI, cs.CV, cs.LG
发布日期: 2026-10-14
备注: 19 pages, 8 figures
💡 一句话要点
提出ActiveEvidence以解决多模态智能体决策问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 多模态智能体 主动证据获取 反事实交互图 决策优化 证书验证
📋 核心要点
- 现有方法在多模态智能体决策中面临如何有效获取视觉证据的挑战,尤其是在成本和决策准确性之间的权衡。
- 论文提出了一种反事实证据交互图(CEIG),通过决策加权封闭目标和差距导向的细化方法来优化证据获取过程。
- 在640个案例的实验中,控制器达到了0.842的任务得分和0.873的证据召回率,显示出显著的性能提升。
📝 摘要(中文)
多模态智能体必须决定是否值得获取视觉证据、查询哪个候选项或分辨率,以及何时进一步交互不再改变其决策。本文在严格的免费元数据/付费观察边界下,研究了这一问题,并将其视为可决策的主动证据获取。提出了反事实证据交互图(CEIG),表示一元证据值、签名的二阶交互和有向条件损害,同时保持未解决交互的显式得分封闭。通过决策加权封闭目标收紧可以改变行动身份的界限,采用差距导向的细化方法,仅评估能够替代现有获取的交互,并在$ε$-决策证书显示未解决项无法超出声明容忍度改变所选行动后停止。实验结果表明,该方法在640个案例中达到了0.842的任务得分和0.816的基础F1值。
🔬 方法详解
问题定义:本文旨在解决多模态智能体在决策过程中如何有效获取视觉证据的问题。现有方法在处理成本和决策准确性之间的平衡时存在不足,尤其是在面对复杂的交互时。
核心思路:论文的核心思路是通过构建反事实证据交互图(CEIG),将一元证据值与二阶交互结合,优化决策过程。通过决策加权封闭目标,论文收紧了可能改变行动身份的界限,从而提高决策的有效性。
技术框架:整体架构包括反事实证据交互图(CEIG)、决策加权封闭目标、差距导向的细化方法和决策证书的验证模块。主要流程为:首先构建CEIG,然后通过差距导向的细化评估交互,最后生成决策证书。
关键创新:最重要的技术创新点在于引入了决策证书的概念,能够在固定的执行轨迹下验证政策一致的最优性。这一方法与现有方法的本质区别在于其能够在不增加额外成本的情况下,优化决策过程。
关键设计:关键设计包括决策加权封闭目标的设置、差距导向细化的实现,以及证书验证的独立实现。这些设计确保了模型的相对性和语义质量的独立评估。实验中使用了640个案例进行训练、验证和测试,确保了方法的有效性。
🖼️ 关键图片
📊 实验亮点
实验结果显示,该控制器在640个案例中达到了0.842的任务得分、0.816的基础F1值和0.873的证据召回率,且在决策过程中仅评估19.6个交互项,相较于191.4个交互项显著降低了计算复杂度,展示了该方法的高效性。
🎯 应用场景
该研究的潜在应用领域包括智能助手、自动驾驶、机器人视觉等多模态交互系统。通过优化证据获取决策,该方法能够提高系统的响应速度和决策准确性,具有重要的实际价值和未来影响。
📄 摘要(原文)
A multimodal agent must decide whether visual evidence is worth acquiring, which candidate or resolution to query, and when further interaction can no longer change its decision. We study this problem under a strict free-metadata/paid-observation boundary, stochastic vector costs, and a frozen execution trace, and treat it as decision-certifiable active evidence acquisition. A Counterfactual Evidence Interaction Graph (CEIG) represents unary evidence value, signed second-order interaction, and directed conditional harm while maintaining explicit score enclosures for unresolved interactions; a decision-weighted enclosure objective tightens the bounds that can change action identity. Gap-directed refinement evaluates only interactions that can displace the incumbent acquisition and stops once an $ε$-decision certificate shows that unresolved terms cannot change the selected action beyond a declared tolerance. The certificate is model-relative: it verifies policy-consistent optimality under the pinned learned score, and semantic quality is measured separately through post-freeze labels. Paired continuation rollouts learn the downstream value of paid previews, an online operational cost filter is separated from independent finite-policy-family trajectory certification, and an acquisition monitor with an independently implemented verifier binds every certificate to the frozen trace. On 640 cases split into 384/128/128 training, validation, and test cases, the controller reaches task score 0.842, grounding F1 0.816, evidence recall 0.873, 1.74 accepted calls, 286.5K visual tokens, and 742ms p95 latency. Dense replay of issued certificates shows 0.992 action agreement and a 0.988 certificate rate while evaluating 19.6 instead of 191.4 expensive interaction terms per decision.