Shanghai-based XTransfer Introduces Self-developed Large Language Model TradePilot

XTransfer, a Chinese B2B cross-border trade payment platform, has introduced TradePilot, a self-developed large language model (LLM) tailored to the foreign trade financial business. The approach intends to improve payment and financial procedures for small and medium-sized firms (SMEs) engaged in worldwide trade.

TradePilot's release coincides with the integration of artificial intelligence (AI) into a variety of businesses, including foreign trade finance. XTransfer created TradePilot to address the sector's expanding complexities, leveraging AI capabilities to improve risk management, customer service and transaction security for B2B SMEs.

The model’s features include advanced natural language processing (NLP) for managing long-context data and multimodal analysis, which enables the model to interpret a variety of information types such as invoices and logistics papers. This increases anti-money laundering (AML) risk control by matching and verifying buyers and sellers.

The model was created  including company data and market demand and underwent multiple rounds of refinement. In June, XTransfer finished training two versions of TradePilot, which apparently surpassed other well-known LLMs, such as GPT-4, in foreign trade financial assessments.

Initial feedback suggests that TradePilot has improved transaction efficiency, decreased expenses and increased payment security for SMEs. It has also improved XTransfer's AI-powered customer support skills, increasing response rates from 13 per cent to 84.2 per cent.

The LLM uses a distributed computing architecture to enable efficient data processing while focussing on data security and privacy. The concept adheres to international and regional rules via encryption, access control and audit methods. XTransfer intends to continue pushing digital developments in the foreign trade market through other technology innovations like TradePilot.