Reinforcement Learning (RL) has shown significant promise in automated portfolio management; however, effectively balancing risk and return remains a central challenge, as many models fail to adapt to dynamically changing market conditions. We propose Meta-controlled Agents for a Risk-aware System (MARS), a novel framework addressing this through a multi-agent, risk-aware approach. MARS replaces monolithic models with a Heterogeneous Agent Ensemble, where each agent's unique risk profile is enforced by a Safety-Critic network to span behaviors from capital preservation to aggressive growth. A high-level Meta-Adaptive Controller (MAC) dynamically orchestrates this ensemble, shifting reliance between conservative and aggressive agents to minimize drawdown during downturns while seizing opportunities in bull markets. This two-tiered structure leverages behavioral diversity rather than explicit feature engineering to ensure a disciplined portfolio robust across market regimes. Experiments on major international indexes confirm that our framework significantly reduces maximum drawdown and volatility while maintaining competitive returns.

When releasing outputs from confidential data, agencies need to balance the analytical usefulness of the released data with the obligation to protect data subjects'

Appendix of "Does enforcing fairness mitigate biases caused by subpopulation shift?" Assume 1. there are only two groups and the set of risk profiles R R Next we provide a proof of Theorem 4.3 under the additional assumption that The risk set R is convex. The unconstrained risk minimizer on unbiased data is algorithmically fair; i.e. FRM problem (4.3) is convex, so (1.2) implies R A sufficient condition for (4.5) is null P In this section, we state and prove a more general verion of Theorem 4.3 that permits continuous discriminative attributes. Let A be a complemented subspace in B. Then A Finally, we review some relevant background on infinite dimensional optimization.

Many instances of algorithmic bias are caused by subpopulation shifts.

Structured electronic health records (EHR) are essential for clinical prediction. While count-based learners continue to perform strongly on such data, no benchmarking has directly compared them against more recent mixture-of-agents LLM pipelines, which have been reported to outperform single LLMs in various NLP tasks. In this study, we evaluated three categories of methodologies for EHR prediction using the EHRSHOT dataset: count-based models built from ontology roll-ups with two time bins, based on LightGBM and the tabular foundation model TabPFN; a pretrained sequential transformer (CLMBR); and a mixture-of-agents pipeline that converts tabular histories to natural-language summaries followed by a text classifier. We assessed eight outcomes using the EHRSHOT dataset. Across the eight evaluation tasks, head-to-head wins were largely split between the count-based and the mixture-of-agents methods. Given their simplicity and interpretability, count-based models remain a strong candidate for structured EHR benchmarking. The source code is available at: https://github.com/cristea-lab/Structured_EHR_Benchmark.

When releasing outputs from confidential data, agencies need to balance the analytical usefulness of the released data with the obligation to protect data subjects'

Large language models (LLMs) are increasingly used for decision-making tasks under uncertainty; however, their risk profiles and how they are influenced by prompting and alignment methods remain underexplored. Existing studies have primarily examined personality prompting or multi-agent interactions, leaving open the question of how post-training influences the risk behavior of LLMs. In this work, we propose a new pipeline for eliciting, steering, and modulating LLMs' risk profiles, drawing on tools from behavioral economics and finance. Using utility-theoretic models, we compare pre-trained, instruction-tuned, and RLHF-aligned LLMs, and find that while instruction-tuned models exhibit behaviors consistent with some standard utility formulations, pre-trained and RLHF-aligned models deviate more from any utility models fitted. We further evaluate modulation strategies, including prompt engineering, in-context learning, and post-training, and show that post-training provides the most stable and effective modulation of risk preference. Our findings provide insights into the risk profiles of different classes and stages of LLMs and demonstrate how post-training modulates these profiles, laying the groundwork for future research on behavioral alignment and risk-aware LLM design.

In this work, we offer a thorough analytical investigation into the role of shared hyperparameters in a hierarchical Bayesian model, examining their impact on information borrowing and posterior inference. Our approach is rooted in a non-asymptotic framework, where observations are drawn from a mixed-effects model, and a Gaussian distribution is assumed for the true effect generator. We consider a nested hierarchical prior distribution model to capture these effects and use the posterior means for Bayesian estimation. To quantify the effect of information borrowing, we propose an integrated risk measure relative to the true data-generating distribution. Our analysis reveals that the Bayes estimator for the model with a deeper hierarchy performs better, provided that the unknown random effects are correlated through a compound symmetric structure. Our work also identifies necessary and sufficient conditions for this model to outperform the one nested within it. We further obtain sufficient conditions when the correlation is perturbed. Our study suggests that the model with a deeper hierarchy tends to outperform the nested model unless the true data-generating distribution favors sufficiently independent groups. These findings have significant implications for Bayesian modeling, and we believe they will be of interest to researchers across a wide range of fields.

This paper studies the interplay between learning algorithms and graph structure for graph neural networks (GNNs). Existing theoretical studies on the learning dynamics of GNNs primarily focus on the convergence rates of learning algorithms under the interpolation regime (noise-free) and offer only a crude connection between these dynamics and the actual graph structure (e.g., maximum degree). This paper aims to bridge this gap by investigating the excessive risk (generalization performance) of learning algorithms in GNNs within the generalization regime (with noise). Specifically, we extend the conventional settings from the learning theory literature to the context of GNNs and examine how graph structure influences the performance of learning algorithms such as stochastic gradient descent (SGD) and Ridge regression. Our study makes several key contributions toward understanding the interplay between graph structure and learning in GNNs. First, we derive the excess risk profiles of SGD and Ridge regression in GNNs and connect these profiles to the graph structure through spectral graph theory. With this established framework, we further explore how different graph structures (regular vs. power-law) impact the performance of these algorithms through comparative analysis. Additionally, we extend our analysis to multi-layer linear GNNs, revealing an increasing non-isotropic effect on the excess risk profile, thereby offering new insights into the over-smoothing issue in GNNs from the perspective of learning algorithms. Our empirical results align with our theoretical predictions, \emph{collectively showcasing a coupling relation among graph structure, GNNs and learning algorithms, and providing insights on GNN algorithm design and selection in practice.}

Appendix of "Does enforcing fairness mitigate biases caused by subpopulation shift?" Assume 1. there are only two groups and the set of risk profiles R R Next we provide a proof of Theorem 4.3 under the additional assumption that The risk set R is convex. The unconstrained risk minimizer on unbiased data is algorithmically fair; i.e. FRM problem (4.3) is convex, so (1.2) implies R A sufficient condition for (4.5) is null P In this section, we state and prove a more general verion of Theorem 4.3 that permits continuous discriminative attributes. Let A be a complemented subspace in B. Then A Finally, we review some relevant background on infinite dimensional optimization.