Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{203552,
        author = {Sahil Bhoye and Vardhak Kore and Purva Vajire and Dr.A. M. Bagade},
        title = {ROT-MAS: A Framework for Resource-Efficient and Trustworthy LLM Agents via Graphlet-Based Optimization and Multi-Agent Verification},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {11361-11370},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=203552},
        abstract = {This study introduces the Resource-Optimized, Trustworthy Multi-Agent System (ROT-MAS), a unique architecture that uses Large Language Models (LLMs) to fully automate data science pipelines. Even though current automated systems have advanced, substantial inefficiencies and a lack of strong auditability are impeding their industrial implementation. ROT-MAS employs a unique three-tiered architecture to overcome these constraints. In the first part, Multi-Agent Debate (MAD), agents assess suggested workflows using a peer-critique process in order to correct mistakes before they are carried out. An LLM-guided Provenance Tracker, the second part, creates a thorough audit trail and provides granularity levels that can be changed to maximize storage effectiveness. Lastly, by anticipating and ending unnecessary model training cycles, the Wasted Computation Mitigation (WCM) module improves environmental sustainability (Green AutoML). According to experimental results, the framework maintains high precision in automated documentation while dramatically reducing crucial errors, minimizing computational waste, and drastically lowering storage overhead.},
        keywords = {Data Provenance, Automated Data Science, Multi-Agent Systems, Green AutoML, LLM Agents, Multi-Agent Debate (MAD), Wasted Computation Mitigation (WCM), Graphlet Prediction.},
        month = {May},
        }