---
格式版本: 2
标题: "Verifiable abstention makes AI leak diagnosis accountable in water distribution networks"
原文链接: "https://arxiv.org/abs/2608.18836"
发布日期: "2026-08-19"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Wed, 19 Aug 2026 11:54:27 UTC (2,308 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-20T16:02:12+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-20T15:56:53+08:00"
入库时间: "2026-08-20T08:02:12.166Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=AI&searchtype=all"
匹配关键词:
  - "AI"
相关厂家:
  []
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 5
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文讨论供水管网泄漏诊断的AI方法，与超节点/AI Rack/机柜级AI基础设施完全无关，不涉及相关部件、架构、供电散热或量产部署。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-20T16:09:01+08:00"
AI主题相关性: 0
AI来源权威性: 5
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
AI摘要: "研究者提出“可验证弃权”机制，将供水管网AI漏水定位转为可问责决策：系统先对数字孪生验证假设，不满足可验证合同就弃权。现场级噪声下行动决策精确率达96%，独立基准中33处泄漏仅对4处行动且全部正确。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:20:02.021Z"
采集批次: "2026年8月20日14点19分32秒"
采集批次ID: "20260820-141932-079"
去重键: "https://arxiv.org/abs/2608.18836"
---

## Computer Science > Artificial Intelligence

## Title:Verifiable abstention makes AI leak diagnosis accountable in water distribution networks

Authors:[Tianwei Mu](https://arxiv.org/search/cs?searchtype=author&query=Mu,+T), [Yue Wang](https://arxiv.org/search/cs?searchtype=author&query=Wang,+Y), [Mingzhe Yuan](https://arxiv.org/search/cs?searchtype=author&query=Yuan,+M), [Manhong Huang](https://arxiv.org/search/cs?searchtype=author&query=Huang,+M), [Wenhong Wang](https://arxiv.org/search/cs?searchtype=author&query=Wang,+W), [Xuerui Yin](https://arxiv.org/search/cs?searchtype=author&query=Yin,+X), [Qing Luo](https://arxiv.org/search/cs?searchtype=author&query=Luo,+Q), [Min Xiao](https://arxiv.org/search/cs?searchtype=author&query=Xiao,+M), [Hui Yang](https://arxiv.org/search/cs?searchtype=author&query=Yang,+H), [Jun Li](https://arxiv.org/search/cs?searchtype=author&query=Li,+J), [Dan Xue](https://arxiv.org/search/cs?searchtype=author&query=Xue,+D)

[View PDF](https://arxiv.org/pdf/2608.18836) [HTML (experimental)](https://arxiv.org/html/2608.18836v1)

> Abstract:Utilities lose a substantial share of treated water to leakage, yet rarely trust artificial-intelligence localizers to dispatch crews: guessing everywhere cannot justify excavation. The gap is accountability, not accuracy: no method proves when it should not act. Here we recast leak localization as decision-making under verifiable abstention. A physics-grounded executor agent falsifies hypotheses (leak, demand, sensor, valve) against a digital twin; an independent supervisor agent, with a large-language-model (LLM) auditor, checks evidence against a code-verifiable contract, then certifies a dispatch, requests evidence or abstains. Under field-grade noise, a 32% forced baseline becomes 96% decision precision on acted events. On an independently generated benchmark it acts on only 4 of 33 leaks, all correct. A 194-event register of audited real leak locations with twin-simulated pressures and flows yields five excavation dispatches, three correct, and 44% survey recovery at full district precision. Accountable abstention offers a defensible route to autonomous water-infrastructure operation.

| Comments: |  |
| --- | --- |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | [arXiv:2608.18836](https://arxiv.org/abs/2608.18836) \[cs.AI\] |
|  | (or [arXiv:2608.18836v1](https://arxiv.org/abs/2608.18836v1) \[cs.AI\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.18836](https://doi.org/10.48550/arXiv.2608.18836) |

## Submission history

From: Tianwei Mu \[[view email](https://arxiv.org/show-email/ca1decb9/2608.18836)\]  
**\[v1\]** Wed, 19 Aug 2026 11:54:27 UTC (2,308 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.18836) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))