CyberSec Research

This paper investigates whether large language models (LLMs) can improve cross-sectional momentum strategies by extracting predictive signals from firm-specific news. We combine daily U.S. equity returns for S&P 500 constituents with high-frequency news data and use prompt-engineered queries to ChatGPT that inform the model when a stock is about to enter a momentum portfolio. The LLM evaluates whether recent news supports a continuation of past returns, producing scores that condition both stock selection and portfolio weights. An LLM-enhanced momentum strategy outperforms a standard long-only momentum benchmark, delivering higher Sharpe and Sortino ratios both in-sample and in a truly out-of-sample period after the model's pre-training cut-off. These gains are robust to transaction costs, prompt design, and portfolio constraints, and are strongest for concentrated, high-conviction portfolios. The results suggest that LLMs can serve as effective real-time interpreters of financial news, adding incremental value to established factor-based investment strategies.

While investment funds publicly disclose their objectives in broad terms, their managers optimize for complex combinations of competing goals that go beyond simple risk-return trade-offs. Traditional approaches attempt to model this through multi-objective utility functions, but face fundamental challenges in specification and parameterization. We propose a generative framework that learns latent representations of fund manager strategies without requiring explicit utility specification. Our approach directly models the conditional probability of a fund's portfolio weights, given stock characteristics, historical returns, previous weights, and a latent variable representing the fund's strategy. Unlike methods based on reinforcement learning or imitation learning, which require specified rewards or labeled expert objectives, our GAN-based architecture learns directly from the joint distribution of observed holdings and market data. We validate our framework on a dataset of 1436 U.S. equity mutual funds. The learned representations successfully capture known investment styles, such as "growth" and "value," while also revealing implicit manager objectives. For instance, we find that while many funds exhibit characteristics of Markowitz-like optimization, they do so with heterogeneous realizations for turnover, concentration, and latent factors. To analyze and interpret the end-to-end model, we develop a series of tests that explain the model, and we show that the benchmark's expert labeling are contained in our model's encoding in a linear interpretable way. Our framework provides a data-driven approach for characterizing investment strategies for applications in market simulation, strategy attribution, and regulatory oversight.

We study the excess growth rate -- a fundamental logarithmic functional arising in portfolio theory -- from the perspective of information theory. We show that the excess growth rate can be connected to the R\'{e}nyi and cross entropies, the Helmholtz free energy, L. Campbell's measure of average code length and large deviations. Our main results consist of three axiomatic characterization theorems of the excess growth rate, in terms of (i) the relative entropy, (ii) the gap in Jensen's inequality, and (iii) the logarithmic divergence that generalizes the Bregman divergence. Furthermore, we study maximization of the excess growth rate and compare it with the growth optimal portfolio. Our results not only provide theoretical justifications of the significance of the excess growth rate, but also establish new connections between information theory and quantitative finance.

We introduce Entropy-Guided Multiplicative Updates (EGMU), a convex optimization framework for constructing multi-factor target-exposure portfolios by minimizing Kullback-Leibler divergence from a benchmark under linear factor constraints. We establish feasibility and uniqueness of strictly positive solutions when the benchmark and targets satisfy convex-hull conditions. We derive the dual concave formulation with explicit gradient, Hessian, and sensitivity expressions, and provide two provably convergent solvers: a damped dual Newton method with global convergence and local quadratic rate, and a KL-projection scheme based on iterative proportional fitting and Bregman-Dykstra projections. We further generalize EGMU to handle elastic targets and robust target sets, and introduce a path-following ordinary differential equation for tracing solution trajectories. Stable and scalable implementations are provided using LogSumExp stabilization, covariance regularization, and half-space KL projections. Our focus is on theory and reproducible algorithms; empirical benchmarking is optional.

In response to growing demand for resilient and transparent financial instruments, we introduce a novel framework for replicating private equity (PE) performance using liquid, AI-enhanced strategies. Despite historically delivering robust returns, private equity's inherent illiquidity and lack of transparency raise significant concerns regarding investor trust and systemic stability, particularly in periods of heightened market volatility. Our method uses advanced graphical models to decode liquid PE proxies and incorporates asymmetric risk adjustments that emulate private equity's unique performance dynamics. The result is a liquid, scalable solution that aligns closely with traditional quarterly PE benchmarks like Cambridge Associates and Preqin. This approach enhances portfolio resilience and contributes to the ongoing discourse on safe asset innovation, supporting market stability and investor confidence.

Recent work has emphasized the diversification benefits of combining trend signals across multiple horizons, with the medium-term window-typically six months to one year-long viewed as the "sweet spot" of trend-following. This paper revisits this conventional view by reallocating exposure dynamically across horizons using a Bayesian optimization framework designed to learn the optimal weights assigned to each trend horizon at the asset level. The common practice of equal weighting implicitly assumes that all assets benefit equally from all horizons; we show that this assumption is both theoretically and empirically suboptimal. We first optimize the horizon-level weights at the asset level to maximize the informativeness of trend signals before applying Bayesian graphical models-with sparsity and turnover control-to allocate dynamically across assets. The key finding is that the medium-term band contributes little incremental performance or diversification once short- and long-term components are included. Removing the 125-day layer improves Sharpe ratios and drawdown efficiency while maintaining benchmark correlation. We then rationalize this outcome through a minimum-variance formulation, showing that the medium-term horizon largely overlaps with its neighboring horizons. The resulting "barbell" structure-combining short- and long-term trends-captures most of the performance while reducing model complexity. This result challenges the common belief that more horizons always improve diversification and suggests that some forms of time-scale diversification may conceal unnecessary redundancy in trend premia.

We study a goal-based portfolio selection problem in which an investor aims to meet multiple financial goals, each with a specific deadline and target amount. Trading the stock incurs a strictly positive transaction cost. Using the stochastic Perron's method, we show that the value function is the unique viscosity solution to a system of quasi-variational inequalities. The existence of an optimal trading strategy and goal funding scheme is established. Numerical results reveal complex optimal trading regions and show that the optimal investment strategy differs substantially from the V-shaped strategy observed in the frictionless case.

We present a multi-agent, AI-driven framework for fundamental investing that integrates macro indicators, industry-level and firm-specific information to construct optimized equity portfolios. The architecture comprises: (i) a Macro agent that dynamically screens and weights sectors based on evolving economic indicators and industry performance; (ii) four firm-level agents -- Fundamental, Technical, Report, and News -- that conduct in-depth analyses of individual firms to ensure both breadth and depth of coverage; (iii) a Portfolio agent that uses reinforcement learning to combine the agent outputs into a unified policy to generate the trading strategy; and (iv) a Risk Control agent that adjusts portfolio positions in response to market volatility. We evaluate the system on the constituents by the CSI 300 Index of China's A-share market and find that it consistently outperforms standard benchmarks and a state-of-the-art multi-agent trading system on risk-adjusted returns and drawdown control. Our core contribution is a hierarchical multi-agent design that links top-down macro screening with bottom-up fundamental analysis, offering a robust and extensible approach to factor-based portfolio construction.

In this work, we aim to develop a market-implied sustainability score for companies, based on the extent to which a stock is over- or under-represented in sustainable funds compared to traditional ones. To identify sustainable funds, we rely on the Sustainable Finance Disclosure Regulation (SFDR), a European framework designed to clearly categorize investment funds into different classes according to their commitment to sustainability. In our analysis, we classify as sustainable those funds categorized as Article 9 - also known as "dark green" - and compare them to funds categorized as Article 8 or Article 6. We compute an SFDR Market-Implied Sustainability (SMIS) score for a large set of European companies. We then conduct an econometric analysis to identify the factors influencing SMIS and compare them with state-of-the-art ESG (Environmental, Social, and Governance) scores provided by Refinitiv. Finally, we assess the realized risk-adjusted performance of stocks using portfolio-tilting strategies. Our results show that SMIS scores deviate substantially from traditional ESG scores and that, over the period 2010-2023, companies with high SMIS have been associated with significant financial outperformance.

Investment style groups investment approaches to predict portfolio return variations. This study examines the relationship between investment style, style consistency, and risk-adjusted returns of Indian equity mutual funds. The methodology involves estimating size and style beta coefficients, identifying breakpoints, analysing investment styles, and assessing risk-shifting intensity. Funds transition across styles over time, reflecting rotation, drift, or strengthening trends. Many Mid Blend funds remain in the same category, while others shift to Large Blend or Mid Value, indicating value-oriented strategies or large-cap exposure. Some funds adopt high-return styles like Small Value and Small Blend, aiming for alpha through small-cap equities. Performance changes following risk structure shifts are analyzed by comparing pre- and post-shift metrics, showing that style adjustments can enhance returns based on market conditions. This study contributes to mutual fund evaluation literature by highlighting the impact of style transitions on returns.

Large Language Models (LLMs) have recently gained popularity in stock trading for their ability to process multimodal financial data. However, most existing methods focus on single-stock trading and lack the capacity to reason over multiple candidates for portfolio construction. Moreover, they typically lack the flexibility to revise their strategies in response to market shifts, limiting their adaptability in real-world trading. To address these challenges, we propose 3S-Trader, a training-free framework that incorporates scoring, strategy, and selection modules for stock portfolio construction. The scoring module summarizes each stock's recent signals into a concise report covering multiple scoring dimensions, enabling efficient comparison across candidates. The strategy module analyzes historical strategies and overall market conditions to iteratively generate an optimized selection strategy. Based on this strategy, the selection module identifies and assembles a portfolio by choosing stocks with higher scores in relevant dimensions. We evaluate our framework across four distinct stock universes, including the Dow Jones Industrial Average (DJIA) constituents and three sector-specific stock sets. Compared with existing multi-LLM frameworks and time-series-based baselines, 3S-Trader achieves the highest accumulated return of 131.83% on DJIA constituents with a Sharpe ratio of 0.31 and Calmar ratio of 11.84, while also delivering consistently strong results across other sectors.

In this paper, we show that interventionally robust optimization problems in causal models are continuous under the $G$-causal Wasserstein distance, but may be discontinuous under the standard Wasserstein distance. This highlights the importance of using generative models that respect the causal structure when augmenting data for such tasks. To this end, we propose a new normalizing flow architecture that satisfies a universal approximation property for causal structural models and can be efficiently trained to minimize the $G$-causal Wasserstein distance. Empirically, we demonstrate that our model outperforms standard (non-causal) generative models in data augmentation for causal regression and mean-variance portfolio optimization in causal factor models.

Since the COVID-19 pandemic, the number of investors in the Indonesia Stock Exchange has steadily increased, emphasizing the importance of portfolio optimization in balancing risk and return. The classical mean-variance optimization model, while widely applied, depends on historical return and risk estimates that are uncertain and may result in suboptimal portfolios. To address this limitation, robust optimization incorporates uncertainty sets to improve portfolio reliability under market fluctuations. This study constructs such sets using moving-window and bootstrapping methods and applies them to Indonesian banking stock data with varying risk-aversion parameters. The results show that robust optimization with the moving-window method, particularly with a smaller risk-aversion parameter, provides a better risk-return trade-off compared to the bootstrapping approach. These findings highlight the potential of the moving-window method to generate more effective portfolio strategies for risk-tolerant investors.

Quantitative trading strategies rely on accurately ranking stocks to identify profitable investments. Effective portfolio management requires models that can reliably order future stock returns. Transformer models are promising for understanding financial time series, but how different training loss functions affect their ability to rank stocks well is not yet fully understood. Financial markets are challenging due to their changing nature and complex relationships between stocks. Standard loss functions, which aim for simple prediction accuracy, often aren't enough. They don't directly teach models to learn the correct order of stock returns. While many advanced ranking losses exist from fields such as information retrieval, there hasn't been a thorough comparison to see how well they work for ranking financial returns, especially when used with modern Transformer models for stock selection. This paper addresses this gap by systematically evaluating a diverse set of advanced loss functions including pointwise, pairwise, listwise for daily stock return forecasting to facilitate rank-based portfolio selection on S&P 500 data. We focus on assessing how each loss function influences the model's ability to discern profitable relative orderings among assets. Our research contributes a comprehensive benchmark revealing how different loss functions impact a model's ability to learn cross-sectional and temporal patterns crucial for portfolio selection, thereby offering practical guidance for optimizing ranking-based trading strategies.

We present the unified market-based description of returns and variances of the trades with shares of a particular security, of the trades with shares of all securities in the market, and of the trades with the market portfolio. We consider the investor who doesn't trade the shares of his portfolio he collected at time t0 in the past. The investor observes the time series of the current trades with all securities made in the market during the averaging interval. The investor may convert these time series into the time series that model the trades with all securities as the trades with a single security and into the time series that model the trades with the market portfolio as the trades with a single security. That establishes the same description of the returns and variances of the trades with a single security, the trades with all securities in the market, and the market portfolio. We show that the market-based variance, which accounts for the impact of random change of the volumes of consecutive trades with securities, takes the form of Markowitz's (1952) portfolio variance if the volumes of consecutive trades with all market securities are assumed constant. That highlights that Markowitz's (1952) variance ignores the effects of random volumes of consecutive trades. We compare the market-based variances of the market portfolio and of the trades with all market securities, consider the importance of the duration of the averaging interval, and explain the economic obstacles that limit the accuracy of the predictions of the returns and variances at best by Gaussian distributions. The same methods describe the returns and variances of any portfolio and the trades with its securities.

Most financial recommendation systems often fail to account for key behavioral and regulatory factors, leading to advice that is misaligned with user preferences, difficult to interpret, or unlikely to be followed. We present FLARKO (Financial Language-model for Asset Recommendation with Knowledge-graph Optimization), a novel framework that integrates Large Language Models (LLMs), Knowledge Graphs (KGs), and Kahneman-Tversky Optimization (KTO) to generate asset recommendations that are both profitable and behaviorally aligned. FLARKO encodes users' transaction histories and asset trends as structured KGs, providing interpretable and controllable context for the LLM. To demonstrate the adaptability of our approach, we develop and evaluate both a centralized architecture (CenFLARKO) and a federated variant (FedFLARKO). To our knowledge, this is the first demonstration of combining KTO for fine-tuning of LLMs for financial asset recommendation. We also present the first use of structured KGs to ground LLM reasoning over behavioral financial data in a federated learning (FL) setting. Evaluated on the FAR-Trans dataset, FLARKO consistently outperforms state-of-the-art recommendation baselines on behavioral alignment and joint profitability, while remaining interpretable and resource-efficient.

In this work, we combine the mean-field game theory with the classical idea of binomial tree framework, pioneered by Sharpe and Cox, Ross & Rubinstein, to solve the equilibrium price formation problem for the stock. For agents with exponential utilities and recursive utilities of exponential type, we prove the existence of a unique mean-field equilibrium and derive an explicit formula for equilibrium transition probabilities of the stock price by restricting its trajectories onto a binomial tree. The agents are subject to stochastic terminal liabilities and incremental endowments, both of which are dependent on unhedgeable common and idiosyncratic factors, in addition to the stock price path. Finally, we provide numerical examples to illustrate the qualitative effects of these components on the equilibrium price distribution.

The empirical results have shown that firstly, with one-week holding period and reinvesting, for SSE Composite Index stocks, the highest p-ratio investment strategy produces the largest annualized rate of return; and for NYSE Composite Index stocks, all the three strategies with both one-week and one-month periods generate negative returns. Secondly, with non-reinvesting, for SSE Composite Index stocks, the highest p-ratio strategy with one-week holding period yields the largest annualized rate of return; and for NYSE Composite stocks, the one-week EEF strategy produces a medium annualized return. Thirdly, under the one-week EEF investment strategy, for NYSE Composite Index stocks, the right frontier yields a higher annualized return, but for SSE Composite Index stocks, the left frontier (stocks on the empirical efficient frontier) yields a higher annualized return than the right frontier. Fourthly, for NYSE Composite Index stocks, there is a positive linear relationship between monthly return and the p-index, but no such relationship is evident for SSE Composite Index stocks. Fifthly, for NYSE Composite Index stocks, the traditional five-factor model performs poorly, and adding the p-index as a sixth factor provides incremental information.