Two-Stage Compliance Detection for Power Enterprises Based on NLI and LLM

Enterprise compliance refers to ensuring that a company's operations adhere to relevant laws, regulations, industry standards, and internal policies. Non-compliance can result in legal liabilities, financial losses, and reputational damage, making it a critical issue in corporate governance. With the advancement of generative artificial intelligence, particularly Large Language Models (LLM), the potential for automated compliance analysis is emerging, offering new methods and tools for enterprise compliance detection. This paper proposes a two-stage compliance detection framework that combines Natural Language Inference (NLI) models and LLM to enhance the efficiency and interpretability of compliance detection between corporate policies and higher-level laws. The framework leverages NLI models for efficient reasoning in detecting semantic conflicts, while LLM are used to generate natural language explanations for deeper semantic analysis, thereby improving both the accuracy and interpretability of compliance detection. Experimental validation in the power industry demonstrates that the proposed framework significantly enhances detection efficiency and reduces compliance risks.