Ant Group, the Chinese financial technology giant, has unveiled its financial large language model (LLM) at the INCLUSION·Conference on the Bund. The company also introduced two new applications powered by the financial LLM – Zhixiaobao 2.0, an intelligent financial assistant for consumers, and Zhixiaozhu 1.0, an intelligent business assistant for professionals in the financial industry.
According to Fin-Eval, Ant Group’s financial LLM outperforms general-purpose LLMs in key areas such as cognition, generation, domain knowledge, professional thinking, and compliance. In 28 financial-specific task categories, the financial LLM also performed better than the industry average in tests including insights extraction, understanding financial intentions, and reasoning on financial events.
To achieve its superior performance, Ant Group trained its financial LLM on hundreds of billions of token datasets containing Chinese financial documents and over 1,000 billion tokens from general corpus datasets. Additionally, the financial LLM utilizes a dataset of more than 600,000 instructions from over 300 real-world industry use cases. This extensive training has significantly optimized the model’s capability for financial-specific tasks. The financial LLM was fine-tuned based on Ant Group’s self-developed general-purpose LLM, which features high efficiency in computing power use.
With its technical capability, the financial LLM can power a wide range of professional services in wealth management, market analysis, investor education, insurance product explanation, family insurance plans, and insurance claims verification.
Wang Xiaohang, Vice President of Ant Group and head of financial LLM, emphasized the importance of domain knowledge and expertise in the development of financial LLMs. Ant Group has integrated LLM technologies with domain knowledge and industry services based on its extensive experiences in financial service businesses.
Ant Group plans to leverage the power of its financial LLM in collaboration with institutions across all digital financial services in the Chinese market.
In addition to the financial LLM, Ant Group also open sourced its AI-powered coding platform CodeFuse for the developer community. The company also made its financial-specific AI task benchmark Fin-Eval available to the public to facilitate the innovation of LLM for the financial industry.
The two new applications powered by the financial LLM, Zhixiaobao 2.0 and Zhixiaozhu 1.0, offer high-quality services to consumers and financial professionals, respectively. Zhixiaobao 2.0 provides market analysis, portfolio diagnosis, asset allocation suggestions, and investor education, achieving a 95% accuracy rate in financial intention recognition. Zhixiaozhu 1.0 supports tasks such as investment analysis, information extraction, content creation, business opportunity insights, and financial tool utilization.
Zhixiaobao 2.0 has been in closed testing for about six months and will be made available to users after regulatory approval, while Zhixiaozhu 1.0 is currently undergoing additional closed testing on Ant Group’s financial platforms.
Ant Group began the financial LLM project in late 2022 and is also exploring other AI technologies like knowledge graph, operations optimization, graph learning, trustworthy AI, and LLM to support the development of inclusive financial services.
Follow crowdfundingmagazine on Instagram: @crowdfundingmagazine_it