Analytics-Everywhere-Lab

Consensus-based multi-agent LLM framework for HTS tariff code classification in maritime logistics

Researchers propose a multi-agent LLM framework for classifying Canadian 10-digit Harmonized Tariff Schedule codes, targeting smart-port and maritime logistics applications. The system combines semantic retrieval over official tariff documents, consensus-based validation, element-wise voting across hierarchical code components, and human-in-the-loop escalation. Evaluated on 3,300 expert-labeled product records, results show exact 10-digit classification remains difficult even for advanced LLMs, with accuracy degrading at finer granularity levels. The work argues for uncertainty-aware, evidence-grounded workflows over fully autonomous classification.