Distill-Belief: Closed-Loop Inverse Source Localization and Characterization in Physical Fields

Distill-Belief：物理场中的闭环逆源定位与表征

Abstract:

Closed-loop inverse source localization and characterization (ISLC) requires a mobile agent to select measurements that localize sources and infer latent field parameters under strict time constraints. 闭环逆源定位与表征（ISLC）要求移动智能体在严格的时间限制下，通过选择测量点来定位源并推断潜在的场参数。

The core challenge lies in the belief-space objective: valid uncertainty estimation requires expensive Bayesian inference, whereas using fast learned belief model leads to reward hacking, in which the policy exploits approximation errors rather than actually reducing uncertainty. 其核心挑战在于信念空间的目标设定：有效的不确定性估计需要昂贵的贝叶斯推理，而使用快速的已学习信念模型则会导致“奖励劫持”（reward hacking），即策略会利用近似误差而非真正降低不确定性。

We propose Distill-Belief, a teacher–student framework that decouples correctness from efficiency. A Bayes-correct particle-filter teacher maintains the posterior and supplies a dense information-gain signal, while a compact student distills the posterior into belief statistics for control and an uncertainty certificate for stopping. At deployment, only the student is used, yielding constant per-step cost. 我们提出了 Distill-Belief，这是一个将正确性与效率解耦的教师-学生框架。一个符合贝叶斯准则的粒子滤波器教师负责维护后验概率并提供密集的信息增益信号，而一个紧凑的学生模型则将后验概率提炼为用于控制的信念统计量以及用于停止决策的不确定性证书。在部署阶段，仅使用学生模型，从而实现恒定的单步计算成本。

Experiments on seven field modalities and two stress tests show that Distill-Belief consistently reduces sensing cost and improves success, posterior contraction, and estimation accuracy over baselines, while mitigating reward hacking. 在七种场模态和两项压力测试上的实验表明，与基准方法相比，Distill-Belief 在降低感知成本、提高成功率、后验收缩和估计精度方面表现一致，同时有效缓解了奖励劫持问题。

---

Paper Details:

• Authors: Yiwei Shi, Zixing Song, Mengyue Yang, Cunjia Liu, Weiru Liu
• arXiv ID: 2604.26095
• Subject: Artificial Intelligence (cs.AI)
• Submission Date: 28 Apr 2026

论文详情：

• 作者： Yiwei Shi, Zixing Song, Mengyue Yang, Cunjia Liu, Weiru Liu
• arXiv ID: 2604.26095
• 学科分类： 人工智能 (cs.AI)
• 提交日期： 2026年4月28日