Researchers have developed Taxon, a novel framework designed to improve the accuracy of hierarchical tax code prediction for e-commerce platforms. This system utilizes a mixture-of-experts architecture combined with guidance from large language models acting as domain experts to ensure semantic alignment between product titles and tax definitions. Taxon has been successfully deployed within Alibaba's tax service system, processing hundreds of thousands of daily queries and demonstrating enhanced accuracy and robustness. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Enhances automation and accuracy in e-commerce compliance, potentially reducing financial inconsistencies and regulatory risks.
RANK_REASON This is a research paper detailing a new framework for tax code prediction.