SHANGHAI–(BUSINESS WIRE)–Ant Group today unveiled its financial large language model (“the financial LLM”) at the 2023 INCLUSION·Conference on the Bund, alongside two new applications powered by the financial LLM including Zhixiaobao 2.0, an intelligent financial assistant for consumers, and Zhixiaozhu 1.0, an intelligent business assistant serving financial industry professionals.
According to Fin-Eval, Ant Group’s in-house benchmark developed to measure the performance of AI in finance, the company’s financial LLM outperforms general-purpose LLMs across 5 key areas, including cognition, generation, domain knowledge, professional thinking, and compliance. In Fin-Eval’s 28 financial-specific task categories, Ant Group’s financial LLM also outperformed industry average in tests such as insights extraction on researches and analysis, understanding of financial intentions and reasoning on financial events.
To achieve this performance, Ant Group’s financial LLM is trained on over hundreds of billions of token datasets containing Chinese financial documents, in addition to over 1,000 billion tokens from general corpus datasets. Furthermore, the financial LLM also utilizes a dataset of more than 600,000 instructions from over 300 real-world industry use cases, significantly optimizing its 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. For example, Ant Group’s general-purpose model utilizes a heterogeneous hardware cluster, which supports up to ten thousand GPUs of different kinds, with its model FLOPs utilization (MFU) for a thousand GPUs reaching 40%. Moreover, the model’s Reinforcement Learning from Human Feedback (RLHF) training throughput under consistent model performance is enhanced by a factor of 3.59 and inference performance is improved by approximately 2 times compared to the existing industry benchmark.
With its technical capability, the financial LLM is able to power a wide range of professional services, encompassing wealth management such as financial product evaluation, market analysis, and investor education, as well as insurance services such as explaining insurance products, making family insurance plans, and verifying insurance claims.
“General-purpose LLMs are difficult to be applied to industry specific use. To truly bring value to the finance industry, language capabilities, domain knowledge, domain expertise and security capabilities are the prerequisite conditions, which are also key challenges to the development of financial LLMs,” said Wang Xiaohang, Vice President of Ant Group and head of financial LLM. “Based on our extensive experiences from financial service businesses, Ant Group has developed an architecture integrating LLM technologies with domain knowledges and industry services, which has already been tested in Ant Groups’ intelligent financial services scenarios.”
Looking ahead, Ant Group will leverage the power of its financial LLM across all digital financial services in collaboration with institutions in China market.
On the same day, 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 industrial innovation of LLM for the financial industry.
Two new applications: Zhixiaobao 2.0 and Zhixiaozhu 1.0
Ant Group’s financial LLM is now in closed tests on its wealth management and insurance platforms, and has been integrated into its various intelligent services.
With the help of LLM technology, the intelligent financial assistant Zhixiaobao 2.0 can provide high quality services to consumers, including market analysis, portfolio diagnosis, asset allocation suggestions, and investors education. It has achieved a 95% accuracy rate in financial intention recognition, with a capability of market analysis and reasoning reaching the averages of professionals, and can engage in high-quality multi-round conversations.
Meanwhile, the intelligent business assistant, Zhixiaozhu 1.0, can be tailored to help various financial professionals such as service representatives, investment advisors and insurance claim specialists, offering comprehensive and intelligent support for tasks such as investment analysis, information extraction, content creation, business opportunity insights, and financial tool utilization.
So far, Zhixiaobao 2.0 has been in closed test for about six months and will be available to users after it receives regulatory approval. Zhixiaozhu 1.0 is currently undergoing additional closed tests by Ant Group and industry partners on the company’s financial platforms.
Ant Group started the financial LLM project at the end of 2022. To better support the development of inclusive financial services, Ant Group is exploring a set of comprehensive AI technologies including knowledge graph, operations optimization, graph learning, trustworthy AI, and LLM.
About Ant Group
Ant Group aims to build the infrastructure and platforms to support the digital transformation of the service industry. Through continuous innovation, we strive to provide all consumers and small and micro businesses equal access to digital financial and other daily life services that are convenient, sustainable and inclusive.
For more information, please visit our website at www.antgroup.com or follow us on Twitter @AntGroup.
Contacts
Media Inquiries
Yinan Duan
Ant Group
duanyinan.dyn@antgroup.com
HOHHOT, CHINA - Media OutReach Newswire - 23 December 2024 - China's National Forestry and…
HONG KONG SAR - EQS Newswire - 23 December 2024 - Novautek Autonomous Driving Limited…
JAKARTA, INDONESIA - Media OutReach Newswire - 23 December 2024 - V-GREEN and PT Xanh…
Increase in opportunities predicted for high-speed optical transceivers and miniaturized connectivity solutions to address data-intensive…
HONG KONG SAR - Media OutReach Newswire - 23 December 2024 - DYXnet, a wholly-owned…
CAIRO, EGYPT - Media OutReach Newswire - 23 December 2024 - Eand, Novo Nordisk, BSH,…