Recent advances in reinforcement learning (RL) have spurred growing interest in its application to multi-period financial planning. Existing literature broadly follows three paradigms: hybrid RL, scalable RL, and end-to-end RL. This paper develops a hybrid regime-aware RL framework for dynamic asset allocation that embeds financial regime structure directly into the RL learning process. A dual-regime model and its corresponding regime-dependent asset sets, proposed in prior work, serve as upstream signals for a downstream RL allocation layer. This regime-aware RL framework introduces six key RL-design innovations: (1) dual-regime forecasts enter the RL state representation and action constraints; (2) a mixture-of-experts architecture assigns separate Bull and Bear agents to different global regimes; (3) an action-masking mechanism restricts each agent to its regime-dependent asset set; (4) a reward structure balances risk and return; (5) Recurrent PPO with LSTM-based actor-critic networks capture temporal dependence and partial observability; and (6) an "Offline-Sim-Online-Deployment" RL training procedure combines synthetic and historical data to improve robustness. Empirical results for a multi-asset portfolio over 1990-2025 show that the proposed regime-aware RL framework outperforms static-weight regime-switching benchmarks by learning adaptive tilts toward recent top-performing assets. Overall, the results highlight the value of integrating dual-regime signals and RL within a unified framework for dynamic asset allocation.
This paper revisits the roles of information, security selection, and portfolio construction in active equity management within a Markowitz mean–variance framework. Using data from 2004–2024 across U.S. and global equity markets, the study evaluates regression-based composite models that integrate valuation metrics, earnings forecasts, revisions, breadth, and momentum. Robust estimation methods, including weighted latent root regression (WLRR) and Least Angle Regression (LAR), show that earnings-related variables, particularly composite forecast measures, consistently generate statistically significant return signals. Longstanding enhanced multifactor models deliver higher information coefficients and forecast efficacy than traditional valuation-only approaches. When implemented within mean–variance optimization, these models produce economically meaningful excess returns in MSCI ACWI and Emerging Markets portfolios, even after accounting for transaction costs and data-mining adjustments. Results further indicate that portfolio construction materially amplifies security selection skill, particularly with higher tracking error and active weights. The paper also explores the possibility of improving forecasting results using alternative highdimensional and machine learning methods and finds that robust regression approaches (WLRR and LAR) deliver the highest out-of-sample information coefficients. These findings, along with related published research, support the notion that disciplined, interpretable methods remain highly competitive tools for equity forecasting during much of the last 30 years. Finally, the analysis shows that expanding the investment universe to include a more comprehensive set of global stocks with earnings forecasts improves Sharpe and information ratios, underscoring the benefits of broad, information-rich opportunity sets and robust statistical modeling in active management.
The advancement of large language models (LLMs) has accelerated the development of autonomous financial trading systems. While mainstream approaches deploy multi-agent systems mimicking analyst and manager roles, they often rely on abstract instructions that overlook the intricacies of real-world workflows, which can lead to degraded inference performance and less transparent decision-making. Therefore, we propose a multi-agent LLM trading framework that explicitly decomposes investment analysis into fine-grained tasks, rather than providing coarse-grained instructions. We evaluate the proposed framework using Japanese stock data, including prices, financial statements, news, and macro information, under a leakage-controlled backtesting setting. Experimental results show that fine-grained task decomposition significantly improves risk-adjusted returns compared to conventional coarse-grained designs. Crucially, further analysis of intermediate agent outputs suggests that alignment between analytical outputs and downstream decision preferences is a critical driver of system performance. Moreover, we conduct standard portfolio optimization, exploiting low correlation with the stock index and the variance of each system's output. This approach achieves superior performance. These findings contribute to the design of agent structure and task configuration when applying LLM agents to trading systems in practical settings.
