On 1 November 2026, Zhongjie Wang and five coauthors submitted Trustworthy Execution of Industrial Agents to arXiv. The preprint proposes a framework within a single enterprise governance domain for agents that modify persistent engineering objects. Its abstract argues that authentication and successful tool calls cannot establish whether intended changes occurred, conclusions have sufficient evidence, or historical results remain applicable.[1]

The authors describe eight control domains spanning authority, planning, composition, tool execution, actual effects, evidence, baseline and use governance, and recovery and evolution. A four-plane architecture and three control chains connect these responsibilities through typed objects and version-bound interfaces. The specification distinguishes tool returns from engineering effects, evidence coverage from engineering judgment, and historical lifecycle state from current applicability.[1]

The abstract reports that an OpenFOAM fixture achieved COMPLETE coverage with a FAIL judgment, while an expired job required readmission despite completed computation. CAD and recovery counterexamples exposed omitted mandatory checks, discarded negative facts, and inconsistent producer-consumer mappings. The authors say targeted repairs support local regression conclusions; their reported 0/29 hazardous commits measures closed-set specification conformance, not a safety probability.[1]

[1]: Trustworthy Execution of Industrial Agents. arXiv, 1 November 2026. https://arxiv.org/abs/2611.02300

Sources

  1. Trustworthy Execution of Industrial AgentsarXiv · 1 Nov 2026

Check or correct this page

Verification record

The digest below shows whether the text you are reading matches the published source. It does not show that the reporting is correct. The build computes it from this page's Markdown source and lists it in the site manifest.

bce1f07554f0e13399abfd2ad94c54776934e07ecb0e8e377f737e9d6d41980e