Published June 2, 2026 | Version 1.0.0
Journal article Open

UVMAF: A Universal Vertical Multilingual Agent Framework for Domain-Specific Autonomous Business Operations in Indian SMEs

Description

Here is exactly what to fill in for every field:

Resource Type

  • Publication → Journal Article

Description

Paste this:

This paper introduces UVMAF—the Universal Vertical Multilingual Agent Framework—a novel hierarchical multi-agent architecture for domain-specific autonomous business operations in Indian SMEs. The framework integrates domain-specialised reasoning, real-time multilingual processing across twelve Indian languages, and vertical orchestration logic. UVMAF employs a three-tier agent topology comprising a Strategic Orchestrator Agent (SOA), Domain Expert Agents (DEAs), and Language Execution Agents (LEAs). Evaluated across four SME verticals—retail, agri-commodity trading, healthcare, and textile export compliance—UVMAF achieves 91.4% task completion rate and 89.7% multilingual instruction-following accuracy. DOI: https://doi.org/10.5281/zenodo.20503150

Files

UVMAF_International_Research_Paper DOI.pdf

Files (262.1 kB)

Name Size Download all
md5:a4489fc4a2dd9d0a4f7f06c580d72254
262.1 kB Preview Download

Additional details

Identifiers

DOI
10.5281/zenodo.20503150
Other
https://orcid.org/0009-0005-3643-2942

Dates

Created
2025-01-01
Submitted
2026-04-09
Available
2026-06-02

References

  • [1] Ministry of MSME, Annual Report 2023-24, Government of India, 2024. [2] Wu et al., AutoGen, arXiv:2308.08155, 2023. [3] Hong et al., MetaGPT, arXiv:2308.00352, 2023. [4] Park et al., Generative Agents, ACM UIST, 2023. [5] Devlin et al., BERT, NAACL-HLT, 2019. [6] Conneau et al., XLM-R, ACL, 2020. [7] Kakwani et al., IndicNLPSuite, EMNLP, 2020. [8] Gala et al., IndicTrans2, TMLR, 2023.