The provided text is an academic paper that investigates the ability of a large language model (LLM) to analyze financial statements and predict future earnings. The authors test GPT-4's performance against human analysts and specialized machine learning models. They find that GPT-4 significantly outperforms human analysts in predicting the direction of earnings changes, even without access to narrative information. GPT-4’s performance is also on par with a state-of-the-art artificial neural network (ANN) model. The authors argue that GPT-4's ability to generate narrative insights from the financial data is key to its success, and they find that trading strategies based on GPT-4's predictions yield higher Sharpe ratios and alphas than other models. Ultimately, the paper suggests that LLMs have the potential to play a central role in financial decision-making.
Paper:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4835311&download=yes
Join Robinhood