VeriMAS-A VerificationGated MultiAgent Framework

The multi-agent LLM literature has shown that language models can fill the roles classical manufacturing agents defined; it has not shown that plants can afford to let them. VeriMAS reframes the problem: autonomy is not a property of capable agents but of a system that can bound what capable agents are permitted to do. By making every agent output a proposal, every execution a verified event, and every uncertain case a human decision supported by simulation evidence, the framework converts trust from an assumption into an architecture—and, through its proposed evaluation, into a measurable one. If verification gates prove cheap relative to the failures they intercept, the path from simulated shopfloors to real ones becomes an engineering program rather than a leap of faith.

[1] P. Leitão, “Agent-based distributed manufacturing control: A state-of-the-art survey,” Engineering Applications of Artificial Intelligence, vol. 22, no. 7, pp. 979–991, 2009. [2] S. Yao et al., “ReAct: Synergizing reasoning and acting in language models,” in Proc. International Conference on Learning Representations (ICLR), 2023. [3] P. Lewis et al., “Retrieval-augmented generation for knowledge-intensive NLP tasks,” in Advances in Neural Information Processing Systems (NeurIPS), 2020. [4] M. Breque, L. De Nul, and A. Petridis, “Industry 5.0: Towards a sustainable, human-centric and resilient European industry,” European Commission, Directorate-General for Research and Innovation, Brussels, 2021. [5] “A Large Language Model-based multi-agent manufacturing system for intelligent shopfloor,” arXiv preprint arXiv:2405.16887, 2024. [Online]. Available: https://arxiv.org/abs/2405.16887 * [6] “Large language model-driven multi-agent systems for improving production efficiency and reducing carbon emissions in manufacturing,” Computers & Industrial Engineering, 2025. [Online]. Available: https://www.sciencedirect.com/science/article/abs/pii/S0360835225004450 * [7] “Self-Evolving Multi-Agent Network for Industrial IoT Predictive Maintenance,” arXiv preprint, 2026. [Online]. Available: https://arxiv.org/pdf/2602.16738 * [8] “Foundation-Model-Based Agents in Industrial Automation: Purposes, Capabilities, and Open Challenges,” arXiv preprint, 2026. [Online]. Available: https://arxiv.org/pdf/2605.02592 * [9] “Integrating Large Language Model Agents with Digital Twins for Industrial Autonomous Systems,” arXiv preprint, 2026. [Online]. Available: https://arxiv.org/pdf/2606.20761 * [10] “FactoryLLM: A Safe and Open-Source AI Playground for Evaluating LLMs in Smart Factories,” arXiv preprint, 2026. [Online]. Available: https://arxiv.org/pdf/2606.14119 *