Quantum-enhanced tax revenue via A-challan: ML, LLMs, and QML approaches

Abstract

In an era of digital transformation, governments around the world are exploring cutting-edge technologies to improve the transparency and efficiency of revenue generation. This paper presents a comprehensive comparative analysis of three quantum-enhanced approaches for government revenue system analysis: Quantum Machine Learning (QML), Quantum Large Language Models (QLLM), and Quantum Advanced Models (QAM). Through empirical evaluation on 44 real Bangladesh government A-Challan documents totaling 5,071,301 in revenue, we demonstrate that the Quantum Advanced Model achieves superior performance with a 100% data processing success rate, 28.05% validated ROI, and 95.2% collection efficiency. In contrast, both QML and QLLM approaches exhibit critical limitations including 22.7% failure rates, negative neural network performance (R2 = - 292.934), and data reliability concerns. Our findings establish the Quantum Advanced Model as the optimal solution for real-world government financial analysis applications, offering immediate deployment capability with concrete financial benefits of 1,422,499.93 annually. This research represents the first successful quantum analysis of real government financial data, positioning Bangladesh as a pioneer in quantum government technology.