AlphaSharpe: LLM-Driven Discovery of Robust Risk-Adjusted Metrics

TL;DR

This paper presents AlphaSharpe, an LLM-driven framework that iteratively evolves financial performance metrics to improve robustness, generalization, and predictive power beyond the traditional Sharpe ratio. By coupling few-shot LLM prompts with an evolutionary loop (crossover, mutation, scoring, ranking), the approach generates metrics such as $\alpha_{S1}$–$\alpha_{S4}$ that better correlate with future performance and yield substantial portfolio benefits. Empirical results demonstrate up to 3× improvements in ranking correlations and up to ~+116% in Calmar-adjusted performance over baselines, across time-series cross-validation and stress periods like the COVID-19 crash. The work also introduces the AlphaSharpe Portfolio, which employs inverse-covariance weighting, stability factors, and entropy regularization to further enhance risk-adjusted returns, with open-source code for reproducibility. Overall, AlphaSharpe showcases the potential of LLM-driven discovery to advance financial analytics and support more robust investment decision-making.

Abstract

Financial metrics like the Sharpe ratio are pivotal in evaluating investment performance by balancing risk and return. However, traditional metrics often struggle with robustness and generalization, particularly in dynamic and volatile market conditions. This paper introduces AlphaSharpe, a novel framework leveraging large language models (LLMs) to iteratively evolve and optimize financial metrics to discover enhanced risk-return metrics that outperform traditional approaches in robustness and correlation with future performance metrics by employing iterative crossover, mutation, and evaluation. Key contributions of this work include: (1) a novel use of LLMs to generate and refine financial metrics with implicit domain-specific knowledge, (2) a scoring mechanism to ensure that evolved metrics generalize effectively to unseen data, and (3) an empirical demonstration of 3x predictive power for future risk-returns, and 2x portfolio performance. Experimental results in a real-world dataset highlight the superiority of discovered metrics, making them highly relevant to portfolio managers and financial decision-makers. This framework not only addresses the limitations of existing metrics but also showcases the potential of LLMs in advancing financial analytics, paving the way for informed and robust investment strategies.