portfolio construction

This study addresses the challenge of constructing robust portfolios under parameter estimation error, market non-stationarity, and trading constraints. The authors propose a framework that circumvents the need to estimate expected returns or covariance matrices by using an equally weighted portfolio as a baseline. The approach integrates dynamic asset eligibility, deterministic rebalancing, and bounded multi-factor tilts, adaptively adjusting factor exposures based on prevailing market conditions—such as volatility and liquidity—through cross-sectional ranking and hard exposure limits. By avoiding reliance on traditional optimization procedures, the method achieves high stability, low turnover, and strong implementability while effectively controlling portfolio concentration and fragility, making it well-suited for long-term asset allocation.

This paper addresses the finite-horizon mean-variance portfolio rebalancing problem in high-dimensional settings (number of assets exceeding sample size), explicitly incorporating both proportional and quadratic transaction costs—both nonconvex—within a unified framework for the first time. We propose a solution methodology integrating nonconvex regularized optimization with high-dimensional statistical inference, and establish theoretical guarantees on algorithmic convergence and statistical consistency of parameter estimation. Monte Carlo simulations and empirical studies on S&P 500 and Russell 2000 datasets demonstrate that our approach significantly improves Sharpe ratios and net returns, confirming the critical performance gains from explicit modeling of nonlinear transaction costs. The core contribution lies in overcoming the small-sample, high-dimensionality barrier to enable joint modeling and provably optimal optimization of realistic, nonconvex transaction cost structures.

This paper addresses two key limitations in multi-asset active allocation: insufficient integration of momentum and trend signals, and weak tail-risk control. To this end, we propose a dual-layer协同 framework that jointly generates and fuses trend signals at both the asset-class level (equities, bonds, commodities) and the risk-factor level (e.g., value, momentum, volatility), embedding them directly into portfolio optimization. Methodologically, the approach combines rolling-window momentum ranking, multi-horizon trend filtering, risk-parity weighting, and volatility-targeting constraints. Its primary contribution lies in the first systematic, cross-dimensional co-modeling of trend signals across assets and factors—simultaneously enhancing returns and mitigating downside risk. A 22-year backtest demonstrates that the strategy delivers an annualized excess return of 3.2% relative to benchmarks including the Bloomberg Barclays US Aggregate Bond Index and the MSCI ACWI Index, while reducing maximum drawdown by 37%.

Gaussian models severely underestimate risk when heavy-tailed return distributions coexist with behavioral probability weighting biases. Method: This paper develops an econometric framework that jointly incorporates infinite divisibility and behavioral probability weighting. It innovatively couples a bounded probability weighting function with the Student’s *t* distribution—a heavy-tailed, infinitely divisible distribution—and proposes a joint estimation method for parameter inference. Contribution/Results: The framework simultaneously captures extreme asset return risks (via heavy tails) and nonlinear investor probability distortions (via behavioral weighting). Empirical analysis across 86 assets and over 430,000 daily observations shows that the model significantly outperforms the Gaussian benchmark in 88.4% of samples. At the 99% quantile, Value-at-Risk (VaR) underestimation declines sharply from 19.7% to 3.2%. Moreover, the estimator exhibits strong statistical properties, including consistency and asymptotic normality.

To address the poor stability and weak adaptability of deep learning models in quantitative investing, this paper proposes an LLM-driven multi-agent collaborative framework for automated discovery and dynamic ensemble optimization of multimodal (numerical, textual, and chart-based) alpha factors. The method innovatively integrates large language models (LLMs), multi-agent systems, and a dynamic weight gating mechanism to establish a market-state-aware, adaptive strategy generation paradigm. Empirically evaluated on the Chinese A-share market, the framework significantly outperforms state-of-the-art baselines: it achieves a 23.6% improvement in Sharpe ratio and a 31.2% reduction in maximum drawdown, effectively balancing return enhancement and risk control. By enabling interpretable, robust, and adaptive decision-making, the proposed framework establishes a novel paradigm for AI-powered quantitative investment.

This study addresses the limitations of traditional small-cap equity strategies, which often fail to effectively integrate news sentiment, macroeconomic indicators, and technical signals while neglecting prediction uncertainty. To overcome these challenges, the authors propose an uncertainty-aware portfolio optimization framework for small-cap stocks. The approach leverages GPT-4o mini to extract sentiment signals from financial news, combines them with macro and technical factors, and explicitly incorporates both aleatoric and epistemic uncertainties from the large language model into the risk covariance matrix. Innovatively, the framework disentangles pure alpha (firm-specific) and pure beta (macro-driven) stock selection mechanisms to enhance interpretability and robustness. Empirical results show that, under a 40-day holding period and 100 basis points of transaction costs, the pure beta strategy achieves a Sharpe ratio of 2.33, and both pure alpha and pure beta strategies consistently outperform hybrid approaches, validating the efficacy of mechanism separation.

This study addresses the limitations of existing risk models in capturing shifts in market regimes and transient factors, which often lead to omitted components in the estimation of asset return covariances. The authors propose an extension to factor models based on maximum likelihood estimation that robustly identifies such transient structures overlooked by the original model, requiring only the realized return series and two hyperparameters—the number of additional factors and a half-life parameter—even in the presence of missing data. By incorporating an exponentially weighted log-likelihood function, the method effectively enhances third-party risk models. Empirical evaluation on the Barra Short-Term US Risk Model demonstrates that the proposed approach significantly improves risk modeling accuracy and successfully recovers covariance structures in returns unexplained by the baseline model.

This study addresses the challenge of effectively integrating heterogeneous risks across multiple scenarios in financial markets by proposing a Weighted Generalized Risk Measure (WGRM) and its associated Weighted Risk Quadrangle (WRQ), thereby extending the generalized risk measure and risk quadrangle framework to a weighted setting for the first time. Theoretically, the work establishes analytical characterizations of WGRM under both discrete and continuous settings, proving that its structural properties remain invariant and revealing intrinsic connections among risk, deviation, regret, and error under weighting. Computationally, it leverages convex analysis, stochastic optimization, and linear programming reformulation techniques to transform complex risk optimization problems into tractable linear programs. Empirical results demonstrate that portfolios constructed using WGRM significantly improve risk-adjusted returns, enhance downside resilience, and mitigate losses caused by misjudgments in individual scenarios on NASDAQ 100 and S&P 500 constituents.

This study investigates the efficient optimization and tail risk measurement of heterogeneous actively managed ETF portfolios. Leveraging daily data from 30 actively managed ETFs and one fixed-income mutual fund, it systematically evaluates static and dynamic strategies—including mean-variance optimization, CVaR minimization, tangency portfolios, and extreme value theory approaches (Hill estimator and Peaks-Over-Threshold with Generalized Pareto Distribution)—under varying constraints that incorporate dependence structures, dynamic allocation, transaction costs, and multidimensional tail risk metrics. The findings indicate that the tangency portfolio delivers superior cumulative returns and risk-adjusted performance, while a dynamic long-only CVaR-95 strategy proves robustly effective. Despite aggregation, portfolios exhibit pronounced downside tail risk. Innovatively treating actively managed ETFs as a joint opportunity set, this work elucidates how strategy heterogeneity collectively shapes overall portfolio performance.

This paper addresses the challenge of jointly optimizing tracking error, portfolio cardinality, and turnover rate in sparse index tracking. We propose a Bayesian sparse modeling framework that integrates uncertainty quantification with implementability constraints. Methodologically, we pioneer the combination of empirical Bayesian stochastic approximation with budget-constrained proximal Langevin Monte Carlo sampling to efficiently approximate high-dimensional posterior weight distributions. We further design an interpretable and tunable rebalancing rule based on posterior activation probabilities and magnitude thresholds. Our contribution lies in unifying sparsity enforcement, dynamic portfolio adjustment, and probabilistic uncertainty characterization—enabling joint quantification of constituent count, tracking deviation, and rebalancing risk. Empirical evaluation on S&P 500 tracking demonstrates significant reductions in both tracking error and turnover, precise control over holding count, and support for decision reliability verification.