-
Financial LLM is in closed testing with financial institution partners on Ant Group’s wealth management and insurance platform
-
The company will implement a master’s degree in finance on its digital financial platform business supporting financial institutions in the Chinese market.
SHANGHAI–(BUSINESS WIRE)–Ant Group today released its Financial Large Language Model (“Financial LLM”) at the 2023 Inclusive Bund Conference and released two new applications based on the Financial LLM, including a smart financial assistant ZhiXiaoBao 2.0 is a smart business assistant that serves consumers, and ZhiXiaoZhu 1.0 is a smart business assistant that serves people in the financial industry.
Fin-Eval, an internal benchmark developed by Ant Group to measure the performance of artificial intelligence in finance, said the company’s Finance Master of Laws outperformed general managers in five key areas: cognition, generation, domain knowledge, professional thinking and compliance. master of Law. Among Fin-Eval’s 28 financial-specific task categories, Ant Group’s financial LLM also outperformed the industry average in tests such as research analysis insight extraction, financial intent understanding, and financial event reasoning.
To achieve this performance, Ant Group’s Finance LLM was trained on over hundreds of billions of token datasets containing Chinese financial documents, in addition to over 1 trillion tokens from a common corpus dataset. In addition, the LLM in Finance leverages a data set of more than 600,000 instructions from more than 300 real industry use cases, significantly optimizing its ability to perform finance-specific tasks.
Financial LLM is fine-tuned on the basis of Ant Group’s self-developed general-purpose LLM, and has the characteristics of high efficiency in the use of computing power. For example, Ant Group’s general model uses heterogeneous hardware clusters to support up to 10,000 GPUs of different types, and the FLOPs utilization rate (MFU) of a model with 1,000 GPUs reaches 40%. In addition, when the model performance is consistent, the model’s human feedback reinforcement learning (RLHF) training throughput is 3.59 times higher than the existing industry benchmark, and the inference performance is improved by about 2 times.
Thanks to its technical capabilities, the Master of Laws in Finance is able to provide a wide range of professional services, including wealth management such as financial product evaluation, market analysis, investor education, and insurance services such as explaining insurance products, developing home insurance plans, and verifying insurance claims.
“A general Master of Laws is difficult to apply to industry-specific purposes. To truly bring value to the financial industry, language ability, domain knowledge, domain expertise and security capabilities are prerequisites and are also key challenges for the development of financial LLM.” Master of Laws in Finance. “Based on our rich experience in the financial services business, Ant Group has developed an architecture that combines LLM technology with domain knowledge and industry services, and has been tested in Ant Group’s smart financial services scenarios.”
Going forward, Ant Group will leverage the power of its Master of Laws in Finance to work with institutions in the Chinese market, covering all digital financial services.
On the same day, Ant Group also open sourced its artificial intelligence-driven coding platform CodeFuse to the developer community. The company also opened Fin-Eval, a financial-specific artificial intelligence task benchmark, to the public to promote industrial innovation in LLM in the financial industry.
Two new applications: Zhixiaobao 2.0 and Zhixiaozhu 1.0
Ant Group Financial LLM is currently in closed testing on its wealth management and insurance platforms and has been integrated into its various smart services.
With the help of LLM technology, the intelligent financial management assistant Zhixiaobao 2.0 can provide consumers with high-quality services including market analysis, investment portfolio diagnosis, asset allocation suggestions, investor education, etc. The financial intention recognition accuracy reaches 95%, the market analysis and reasoning ability reaches the average level of professionals, and it can conduct high-quality multi-round conversations.
At the same time, the intelligent business assistant Zhixiaozhu 1.0 can be customized to help various financial professionals such as service representatives, investment consultants, and insurance claims specialists, providing comprehensive intelligent support for tasks such as investment analysis, information extraction, content creation, and business opportunities. Insights and utilization of financial instruments.
Up to now, Zhixiaobao 2.0 has been in closed testing for about 6 months and will be open to users after receiving regulatory approval. Currently, Smart Piggy 1.0 is undergoing additional closed testing on the financial platform of Ant Group and industry partners.
Ant Group launched the financial LLM project at the end of 2022. In order to better support the development of inclusive financial services, Ant Group is exploring a complete set of comprehensive artificial intelligence technologies including knowledge graph, operation optimization, graph learning, trusted AI, and LLM.
About Ant Group
Ant Group aims to build infrastructure and platforms that support the digital transformation of the service industry. Through continuous innovation, we strive to provide all consumers and small and micro enterprises with equal access to convenient, sustainable, and inclusive digital finance and other daily life services.
For more information, please visit our website at www.antgroup.com or follow us on Twitter @AntGroup.
Contact information
Media inquiries
Yinan Duan
Ant Group
duanyinan.dyn@antgroup.com
source link
Content provided by Business Wire. Jinri Toutiao is not responsible for the content provided or any link related to the content. Toutiao is not responsible for the correctness, topicality or quality of the content.
