BloombergGPT

About

BloombergGPT is Bloomberg’s groundbreaking large language model (LLM), a 50-billion-parameter generative AI specifically designed for financial applications. Released in 2023, it leverages advanced natural language processing (NLP) to outperform similarly sized models on financial tasks while maintaining competitive performance on general-purpose benchmarks. This domain-specific model is trained on an extensive dataset of financial documents spanning 40 years, combined with general-purpose datasets, resulting in a training corpus of over 700 billion tokens. BloombergGPT supports a wide array of NLP tasks, including sentiment analysis, named entity recognition, news classification, and question answering, enabling improved workflows and innovative use cases for Bloomberg’s customers. Bloomberg’s approach combines a curated financial dataset with public data, creating a model fine-tuned for the complexity of financial language while retaining general NLP capabilities. By using this hybrid training method, BloombergGPT excels in finance-specific benchmarks and broader general-purpose tests like BIG-bench Hard and Knowledge Assessments. This innovation showcases Bloomberg’s expertise in integrating machine learning, AI, and NLP into finance, offering faster time-to-market solutions and higher out-of-the-box performance for various applications, from conversational systems to text generation. The development of BloombergGPT exemplifies Bloomberg’s four-decade legacy of financial data curation and innovation. Built with collaboration between its ML Product and Research group and AI Engineering team, the model marks a significant advancement in financial technology. BloombergGPT aims to unlock new opportunities by harnessing the vast data available on the Bloomberg Terminal, offering unparalleled value to the financial domain while setting a new standard for domain-specific AI applications.