Integrating LLM-Generated Views into Mean-Variance Optimization Using the Black-Litterman Model

2025-04-19

AI Summary

Traditional mean-variance optimization suffers from high sensitivity to expected returns, while the Black–Litterman (BL) framework faces challenges in quantifying and calibrating investor views. To address these limitations, this paper proposes the first integration of large language models (LLMs) into the BL portfolio optimization pipeline: LLMs jointly analyze historical price data and corporate metadata to generate stock-level expected returns accompanied by uncertainty estimates (i.e., predictive variances), serving as calibrated, subjective views for the BL model. A biweekly rolling backtest over June 2024–February 2025 demonstrates that the LLM-augmented portfolio significantly outperforms the S&P 500, equal-weighted, and conventional mean-variance portfolios on risk-adjusted metrics—including the Sharpe ratio. Furthermore, empirical analysis identifies the degree of LLM optimism bias and confidence stability as key performance moderators.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Portfolio optimization faces challenges due to the sensitivity in traditional mean-variance models. The Black-Litterman model mitigates this by integrating investor views, but defining these views remains difficult. This study explores the integration of large language models (LLMs) generated views into portfolio optimization using the Black-Litterman framework. Our method leverages LLMs to estimate expected stock returns from historical prices and company metadata, incorporating uncertainty through the variance in predictions. We conduct a backtest of the LLM-optimized portfolios from June 2024 to February 2025, rebalancing biweekly using the previous two weeks of price data. As baselines, we compare against the S&P 500, an equal-weighted portfolio, and a traditional mean-variance optimized portfolio constructed using the same set of stocks. Empirical results suggest that different LLMs exhibit varying levels of predictive optimism and confidence stability, which impact portfolio performance. The source code and data are available at https://github.com/youngandbin/LLM-MVO-BLM.

Problem

• Overcoming sensitivity in traditional mean-variance portfolio optimization models
• Integrating LLM-generated investor views into Black-Litterman framework
• Assessing LLM-based stock return predictions for portfolio performance

Innovation

• LLMs generate stock return predictions
• Black-Litterman integrates LLM views with uncertainty
• Biweekly backtesting compares LLM portfolios to benchmarks