Time Machine for Business and Finance: A New Suite of LLMs Unveiled at AMBS
A new suite of Large Language Models (LLMs) developed by researchers at Alliance 91直播 Business School (AMBS) offers powerful tools for modelling historical business information.
The release of marks an exciting development, introducing the first suite of 68 historical pre-trained Large Language Models (LLMs) specifically designed for business studies. These models function like a time machine, allowing researchers to go back as far as 2007 to analyse historical information.
Developed over more than three years at Alliance 91直播 Business School (AMBS) and the Centre for Financial Technology (FinTech) Studies, these models tackle complex challenges such as look-ahead bias and information leakage, setting a new standard for precision in accounting, finance, and related fields.
This release represents the largest specialised LLM suite to date in terms of the number of models developed. The pre-training process, which spanned a total of three months, underscores the level of effort invested in creating models that offer enhanced reliability for business studies.
Sustainability was a key priority throughout the development of FinText. In alignment with The University of Manchester's broader commitment to sustainability, all electricity used during the pre-training process was fully traceable and sourced exclusively from renewable energy, reinforcing our dedication to environmental responsibility.