Consensus-based Agentic Large Language Model Framework for Harmonized Tariff Schedule Code Classification

The article presents a novel agentic large language model (LLM) framework designed for the classification of Canadian 10-digit Harmonized Tariff Schedule (HTS) codes, addressing challenges posed by ambiguous product descriptions and hierarchical tariff structures. Key features include multi-agent information retrieval, semantic analysis of official documents, evidence-grounded reasoning, and a consensus-based validation approach that incorporates human oversight. Evaluation on a dataset of 3,300 labeled records reveals that while advanced LLMs struggle with fine-grained classification, the proposed framework enhances interpretability and compliance in maritime logistics, underscoring the importance of human-in-the-loop methodologies for accurate HTS classification.