Jul 2026· NLP & Big Data· 0 citations· 54 references
TL;DR
A closed-loop multi-agent decision framework that introduces prompt-level learning as a scalable alternative to full model retraining and highlights the potential of prompt-level adaptation for building robust and autonomous financial decision systems.
Abstract
Traditional portfolio management systems often rely on static rules or fixed prompts, which limits their ability to adapt to changing market conditions. This paper proposes a closed-loop multi-agent decision framework that introduces prompt-level learning as a scalable alternative to full model retraining. The architecture comprises specialised agents for market signal extraction, sentiment analysis, macroeconomic interpretation, risk control, and portfolio construction, all coordinated through the DSPy framework and powered by Llama 3.1 8B.A key contribution is a feedback-driven optimisation mechanism that refines agent prompts using realised trading outcomes without human intervention. Moderate drawdowns trigger incremental prompt updates, while severe drawdowns activate full prompt reconfiguration. Empirical evaluation on a six-year dataset (2015–2020) shows that the system achieves cumulative returns above 80% with improved risk-adjusted performance (Sharpe > 1.5), outperforming a SPY buy-and-hold benchmark, including during the COVID-19 market disruption. Overall, the results highlight the potential of prompt-level adaptation for building robust and autonomous financial decision systems.
When economic structures and market dynamics shift, classic portfolio rebalancing algorithms often suffer from unstable and degraded performance. To improve the return and robustness of portfolio management, we explore reinforcement learning (RL) and propose Scenario-Context Rollout (SCR), a macroeconomics-guided feedback mechanism to produce a distribution of next-day joint returns under potential economic shocks. However, doing so faces new challenges, as history will never tell what would have happened differently. As a result, incorporating scenario-based rewards from rollouts introduces a reward-transition mismatch in temporal-difference (TD) learning, destabilizing RL critic training. We theoretically analyze this problem and show that combining scenario-scored rewards with tape-realized transitions induces a hybrid fixed point. Guided by this analysis, we construct a counterfactual next state using the SCR continuations and augment the critic agent's bootstrap target. Doing so stabilizes the learning and provides a viable bias-variance tradeoff. In out-of-sample evaluations across 31 distinct universes of U.S. equity and ETF portfolios, our method improves Sharpe ratio by up to 76% and reduces maximum drawdown by up to 53% compared with classic and RL-based portfolio rebalancing baselines.
Vanya Priscillia Bendatu, Yao Lu· Proceedings of the 32nd ACM...· 0 citations
Large language models (LLMs) based agents are beginning to participate in portfolio construction and market analysis, where decisions must be justified under evolving information and risk constraints. Current assessment practice, however, remains poorly aligned with this setting: many studies rely on static examinations or report only terminal portfolio returns, while the intermediate evidence, analyst judgments, and execution steps that produced those returns stay largely invisible. We introduce NextFund, an evaluation platform that makes financial-agent behavior observable under live market conditions. The platform couples time-consistent market access, coordinated multi-agent analysis, and persistent logging of the full decision path from observation to trade. Through an interactive Trading Arena, users can compare models across markets, inspect equity curves, and drill from leaderboard outcomes down to individual justifications. We present NextFund on Hong Kong, U.S., and China A-share equities, illustrating how inspectable decision histories enable fairer benchmarking and more actionable diagnosis. Our demo is available at https://paradoox.cn/nextfund/.
Changlun Li, Peixian Ma, Qiqi Duan et al.· 0 citations
Automated equity portfolio management is an important challenge in computational finance where the conventional optimization approaches fail to cope with non-stationary market forces and multifaceted decision-making processes. Deep reinforcement learning (DRL) algorithms have become a natural paradigm of sequential portfolio allocation decisions, providing their adaptive learning ability potentially more successful than traditional methods. This review will look at how automated portfolio manAdvanced Actoragement systems based on DRL have developed over time, beginning with single-agent applications, to more complex multi-agent ensemble systems. Our analysis of algorithmic advances, system architecture styles and technical implementation plans that have influenced this domain between 2018 and 2025. We have done a full search based on PRISMA practice by utilizing various scholarly databases, such as IEEE Xplore, ACM Digital Library, arXiv and Google Scholar. We used a search strategy based on deep reinforcement learning algorithms used in equity portfolio management with particular attention to technical applications and systems design. We obtained descriptive algorithm specifications, network structures, training procedures and performance indicators in 156 peer-reviewed articles. The discussion shows that there were four different evolutionary phases, with Phase I (2018–2020) consisting of simple single-agent models such as Deep Q-Networks (DQN) and REINFORCE with a low Sharpe ratio (0.7–1.0). In Phase II (2020–2022), improved phase II policies overcame phase I (2015–2017) results due to the implementation of novel advanced phase II policy gradient methods such as Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), and Soft Actor-Critic (SAC) with Sharpe ratios of 1.2–1.4. The phase III (2022–2024) was the stage of development of multi-agent systems with cooperative, competitive, and hierarchical structures with a Sharpe ratio of 1.5–1.8. Phase IV (2024–2025) is the next horizon where the ensemble systems will incorporate a number of DRL algorithms and exhibit a Sharpe ratio of 1.8–2.4 and accrue a maximum drawdown of about 8%. The systematic development of single-agent to multi-agent ensemble systems shows great improvements in algorithmic and architecture of DRL-based portfolio management. The identification of the essential system design patterns, algorithmic developmental courses and technical implementation issues are some of the contributions. The research in the future focuses on the importance of explainable AI integration, meta-learning market regime adaptation, and consistent evaluation systems in reproducible research.
Aditi Kumar Rout, U. D. Acharya, Prakash K. Aithal et al.· Discover Computing· 0 citations
Retail participants in Indian index options markets typically apply a small set of habitual strategies regardless of the prevailing market environment, which increases the likelihood of mismatched risk exposure. This paper proposes the Multi-Agent Regime Intelligence Framework (MARIF), a conceptual, rule-based multi-agent architecture that classifies the prevailing market regime for NIFTY 50 and BANKNIFTY by combining five heterogeneous information sources: price-based technical indicators, India VIX volatility levels, NSE option-chain analytics (Open Interest and Put-Call Ratio), Market-Wide Position Limit (MWPL) institutional concentration data, and sentiment extracted from financial news headlines. Five specialised agents produce independent signals that a Coordinator Agent fuses through an explicit weighted-voting rule into a single regime label and an associated options-strategy recommendation, such as a Bull Call Spread, Bear Put Spread, Iron Condor, or Long Straddle. This paper presents the architecture, its grounding in established option-pricing theory, an explicit specification of the fusion rule, and a worked illustrative walk-through of how the framework processes a representative set of inputs. No live backtesting has yet been performed; the framework is presented at the design and specification stage, and a concrete empirical validation plan is proposed as immediate future work. The contribution is an interpretable, extensible architecture unifying signal domains that are normally treated separately in Indian derivatives literature, rather than a claim of validated trading performance.
Deepanshu Lamba, Dr. Neelam Srivastav· Journal of Frontiers in Mult...· 0 citations
Effective portfolio management (PM) is a cornerstone of financial strategy, yet it is often challenged by the uncertainties and the high dimensionality of market data. Traditional PM techniques, whether model-based or model-free, frequently fail to address these complexities, resulting in suboptimal asset allocation. This study introduces an asynchronous hierarchical reinforcement learning (AHRL-PM) framework aimed at enhancing active PM. The proposed AHRL-PM approach employs a two-tiered agent system to navigate market complexities. The first-layer agent selects top-performing stocks monthly based on alpha factors, while the second-layer agent adjusts asset weights using market fundamentals and technical indicators, dynamically rebalancing the portfolio in response to market shifts on a daily basis. We rigorously evaluated our AHRL-PM framework through extensive experiments in six globally representative markets. Performance was assessed using two key metrics: annualized return (AR) and Sharpe ratio (SR). Our model consistently outperformed traditional benchmarks on both metrics across all the markets, demonstrating robustness and versatility in PM. Furthermore, ablation studies comparing the AHRL-PM model to a synchronous model and single-layer reinforcement learning (RL) approaches validated its superior performance in terms of profitability and risk-adjusted returns, while also highlighting significant reductions in training time and appropriate portfolio weight adjustments to respond to market dynamics and uncertainties, underscoring the model's efficiency and practical applicability.
Shuyu Liu, Tianxiang Cui, Yiran Li et al.· IEEE Transactions on Neural...· 0 citations
This study proposes an adaptive portfolio optimization framework that integrates Random Forest(RF)-based return forecasting into a Mean-Variance-Forecast Error (MVF) model, augmented by a Hidden Markov Model (HMM) for market regime identification. Using weekly historical return data from 40 LQ45-listed stocks spanning January 2014 to January 2025, the framework dynamically adjusts portfolio allocations in response to bull and bear market conditions detected by a two-state HMM. The primary methodological contribution lies in addressing the static limitation of conventional MVF under shifting market regimes. Out-of-sample evaluation over a 138-week test period demonstrates that regime-switching MVF achieves Sharpe ratios above 1.30, substantially lower maximum drawdowns than the MVF-only portfolio, and cumulative returns of 291.96%. Bootstrap-validated 95% confidence intervals confirm the statistical robustness of these improvements. Nevertheless, portfolio turnover remains high during active reallocations. These findings indicate that combining machine-learning-based predictive modelling with adaptive, regime-driven allocation enhances portfolio stability, mitigates extreme losses, and improves risk-return efficiency under dynamic emerging-market conditions.
Fadly Ramdhani, D. Saepudin· ZERO Jurnal Sains Matematika...· 0 citations