The rapid development of large language models (LLMs) has renewed interest in agent-based modeling (ABM). However, current LLM-based ABM research faces several key challenges: modeling evolving agent-environment interactions, enabling flexible counterfactual reasoning, and automating simulation workflows for scientific research. In this paper, we propose Eco3S, a socio-economic system simulation framework for economic research and policy analysis that addresses these challenges through three key mechanisms: (1) Co-evolving Environment Design, a bidirectional feedback loop where agents and the environment co-evolve, producing realistic emergent behaviors; (2) Structural Causal Simulation, a structural causal model (SCM)-inspired counterfactual mechanism that allows flexible interventions for diverse causal inference tasks; (3) Simulation-Analysis-Refinement Paradigm, a self-corrective mechanism that iteratively refines experimental designs based on prior simulation results. Experiments on diverse economic scenarios confirm Eco3S's effectiveness in replicating multiple established economic studies (canal decay, origins of governance, and information propagation) and phenomena across domains. Additional results further demonstrate its scalability and generalizability, highlighting the framework's potential for rigorous economic research and policy-making.

Constrained multi-objective optimization problems frequently arise in domains, such as industrial production, supply chain management, and finance. However, despite their widespread relevance, these problems have received limited attention in the simulation optimization literature. This paper considers a multi-objective ranking and selection problem with stochastic constraints, where each alternative has multiple performance measures and is subject to different constraints. Given a fixed simulation budget, the objective is to develop an efficient simulation budget allocation strategy to accurately identify the set of Pareto-optimal and feasible alternatives from a finite set of candidates. To this end, an optimal computing budget allocation (OCBA) problem is formulated to maximize the probability of correctly selecting the Pareto feasible set. Then, a lower bound on the probability of correct selection for the Pareto feasible set is derived, and used to construct an OCBA rule along with a sequential allocation procedure. Numerical experiments and a portfolio selection case study demonstrate the effectiveness of the proposed OCBA rule in significantly improving simulation sampling efficiency.

Risk disclosures play a crucial role in the investment decision-making processes for investors. However, extracting relevant variables from unstructured financial text poses a nontrivial challenge. In this paper, we propose RiskBERT, a large language model (LLM) trained on financial texts and risk knowledge graphs, specifically designed for risk factors extraction. By incorporating both finance and risk knowledge, RiskBERT significantly improves the extraction of risk factors in financial texts. We evaluate RiskBERT's performance on a labeled risk factors dataset comprising 119,153 sentences from 2400 Chinese A-listed companies and compare it against other LLMs and automated text analysis algorithms for risk types' classification. Our findings demonstrate that RiskBERT outperforms alternative models, particularly when the training sample size is limited. Moreover, we uncover that RiskBERT provides risk informativeness estimates in annual reports that are at least 4.5% higher than those derived from other models. These results highlight the value of RiskBERT as a powerful tool for extracting risk factors and enhancing risk analysis in finance and accounting domains.

Information theory has inspired numerous advancements in multi-view learning. Most multi-view methods incorporating information-theoretic principles rely an assumption called multi-view redundancy which states that common information between views is necessary and sufficient for down-stream tasks. This assumption emphasizes the importance of common information for prediction, but inherently ignores the potential of unique information in each view that could be predictive to the task. In this paper, we propose a comprehensive information-theoretic multi-view learning framework named CIML, which discards the assumption of multi-view redundancy. Specifically, CIML considers the potential predictive capabilities of both common and unique information based on information theory. First, the common representation learning maximizes Gács-Körner common information to extract shared features and then compresses this information to learn task-relevant representations based on the Information Bottleneck (IB). For unique representation learning, IB is employed to achieve the most compressed unique representation for each view while simultaneously minimizing the mutual information between unique and common representations, as well as among different unique representations. Importantly, we theoretically prove that the learned joint representation is predictively sufficient for the downstream task. Extensive experimental results have demonstrated the superiority of our model over several state-of-art methods. The code is released on CIML.

Sequential federated learning (SFL) trains models collaboratively across clients in a chain manner. This order shows communication efficiency compared to traditional FL in a parallel manner with a star topology. However, SFL can fail to produce stable training results when clients have significant statistical heterogeneity among their local data distributions. To address these challenges, we propose a novel element-wise model decoupling framework named SeqFedEDT that accelerates SFL training by separating model parameters of each client into a shared subset for global knowledge collaboration and a personalized subset for migrating data heterogeneity. We explore three types of parameter contribution scoring metrics based on gradient, Fisher information, and parameter importance (PI) for personalized parameter selection. In addition, we propose a quantile-based thresholding mechanism to separate shared and personalized subsets and explore the best performance quantile selection in numerical studies. Extensive experiments demonstrate that SeqFedEDT outperforms eight state-of-the-art methods across diverse datasets and heterogeneity scenarios.

Energy poverty creates substantial economic, social, and environmental challenges, particularly in developing regions where access to affordable and sustainable energy remains limited. Despite increasing interest in financial innovation, there is still no consensus on which financing mechanisms are most effective in alleviating energy poverty. This study introduces an AI-driven fuzzy decision support framework that integrates Koch snowflake-based fuzzy sets, fractal geometry, and deep learning to identify the most suitable innovative financial products for mitigating energy poverty. The proposed model employs a Siamese network to determine expert weights by analyzing demographic similarities, applies the simple weight calculation (SIWEC) method for criterion weighting, and ranks alternatives using hybrid fuzzy techniques. The findings indicate that risk management and government support are the most critical criteria for designing effective financial instruments, while sukuk and energy leasing emerge as the optimal financial solutions. This research advances the literature by providing a novel fuzzy modeling approach that enhances uncertainty representation and decision accuracy, thereby offering practical insights for policymakers and investors aiming to promote inclusive and sustainable energy finance.

Peer-to-peer (P2P) energy trading is an industry where energy producers and consumers buy and sell energy directly by using a platform and in most cases, the technology behind P2P energy trading is blockchain. Tracking performance indicators of blockchain-based P2P energy trading is important to evaluate the effectiveness of a project and to help investors manage resources effectively. However, the literature is still scarce, and the risks of strategic decision making are higher. To fill this gap, we present a novel method to prioritize strategies in P2P energy trading. The model combines molecular fuzzy-based cognitive maps with molecular fuzzy ranking. It contributes to the literature by creating a novel method for ranking alternatives across different geometric shapes. So, the testing of reliability of ranking can be done and the accuracy and robustness of the model can be improved. The Q-learning approach also allows expert weights to be calculated objectively which mitigates the subjectivity of the results and helps investors make informed decisions. By providing a reliable framework for strategy development, this study contributes significantly to the literature and thus investors can make well-informed decisions. The results show that blockchain scalability and grid integration are the most critical performance indicators to enhance these projects. Community empowerment through local partnerships for microgrid development is also the most important investment option.

Abstract Artificial intelligence (AI) in finance is commonly reviewed by method, data type, or application domain. These perspectives are essential, but they understate a deeper shift: AI is moving from a predictive tool to a component of human–AI hybrid financial decision systems. This integrative and conceptual review synthesizes literature across finance, management, human–computer interaction (HCI), and AI to examine how humans and AI jointly participate in information acquisition, prediction, recommendation, approval, execution, monitoring, and learning. We argue that the central question is moving from model performance to decision architecture: how authority, oversight, and accountability should be allocated across financial workflows. We show that human–AI complementarity in finance is conditional rather than automatic, depending on task structure, private information, feedback quality, incentives, explanation design, and governance. We also argue that AI-mediated financial decisions are reflexive: they reshape organizational workflows, prices, liquidity, credit allocation, and the future data on which subsequent decisions rely. The review integrates evidence on methods, data, scenarios, explainability, trust, governance, financial large language models (FinLLMs), and agentic finance, and organizes the field around an integrated decision-system framework consisting of five connected constructs—delegation frontier, reliance wedge, decision-useful explainable artificial intelligence (XAI), meaningful oversight, and reflexive AI loop—to support cumulative research on investment, trading, credit, asset management, risk, compliance, and financial regulation.

Earthquakes pose a significant threat to energy systems by causing damage to power plants and transmission networks, which can severely disrupt critical services such as healthcare delivery and industrial production. Ensuring the seismic resilience of energy infrastructure is therefore essential for maintaining social stability and economic continuity. Despite its importance, the existing literature lacks systematic and intelligent frameworks for identifying the most effective strategies to mitigate earthquake-induced power outages. To address this gap, this study proposes a novel hybrid decision-making model that integrates Cantor dust fuzzy sets with artificial intelligence-based expert weighting mechanisms, together with weighted power averaging, entropy-based weighting, logarithmic percentage change-driven objective weighting, evaluation based on distance from the average solution, and alternative ranking using two-step logarithmic normalization techniques. The proposed framework exploits the fractal characteristics of Cantor sets to capture subtle differences in expert judgments, while artificial intelligence enables adaptive and data-driven assignment of expert and criterion weights. This integrated structure enhances the robustness, transparency, and reliability of the prioritization process. The empirical findings reveal that technological improvement and government support are the most influential factors in strengthening energy system resilience, whereas microgrids and the widespread deployment of energy storage systems emerge as the most effective strategies for preventing power outages during and after earthquakes. This study contributes to the intersection of energy resilience and artificial intelligence by offering an advanced, interpretable, and scalable decision-support framework for resilient energy infrastructure planning. In line with Sustainable Development Goal 7, the proposed model supports reliable and sustainable energy access in disaster-prone regions.

Improving boarding efficiency reduces airplane turnaround time and improves passenger experience. Airlines typically assign passengers to a few sequential boarding groups using static seat-based rules. Yet arrivals, seat choices, and luggage are sequential and random, and a static rule ignores the seats earlier passengers have already taken. We propose the first dynamic formulation of boarding group assignment. As each passenger checks in, we observe earlier passengers' seats and groups, the current passenger's seat, and optional luggage information, then assign a group while keeping companions together. We formulate dynamic group assignment as a Markov decision process and solve it with reinforcement learning (RL). The policy uses a convolutional neural network to encode the checked-in seat-assignment state and is trained by proximal policy optimization. The reward balances total boarding time and average individual boarding time. We benchmark the proposed RL policy against three companion-compatible static policies (back-to-front, modified Steffen, and alternating block) in an in-house simulator covering six single- and double-aisle layouts. Back-to-front with optimized group sizes achieves the shortest total boarding time and average individual boarding time among the static benchmarks across all layouts. The dynamic RL policy further outperforms it on both metrics in every layout. On a representative case, the RL policy outperforms the optimal back-to-front by up to 9.8% in total boarding time and 22.8% in average individual time. Sweeping the reward weight yields an approximate Pareto frontier for operator choice. Trained policies remain robust under out-of-distribution operating conditions, including varying load factors, companion sizes, and luggage loads.

Consensus is an important issue in large-scale group decision making (LSGDM). Feedback mechanism in consensus reach process (CRP) can provide adjustment opinion for decision makers (DMs) to improve the efficiency of consensus. However, due to the personal interests of DMs, they may refuse to modify their opinions according to the adjustment opinions. Such non-cooperative behavior will hinder CRP and increase the cost of consensus. Consequently, this paper designs a two-layer feedback mechanism to provide adjustment opinions for DMs with the minimum cost and reaching of both local and global consensus in LSGDM through effective interaction with clusters and DMs. Subsequently, considering the subjective revised opinions and objective adjustment difficulties of DMs, this paper defines five new types of non-cooperative behaviors. Furthermore, to address those non-cooperative behaviors, an incentive factor that can be flexibly changed according to the different types of non-cooperative behaviors is constructed to personally manage non-cooperative behaviors. Finally, a detailed numerical example demonstrates the effectiveness of the proposed method, and the advantages of the proposed method are illustrated through sensitive analyses and comparative analyses.

Enterprise financial risk analysis aims at predicting the future financial risk of enterprises. Due to its wide and significant application, enterprise financial risk analysis has always been the core research topic in the fields of Finance and Management. Based on advanced computer science and artificial intelligence technologies, enterprise risk analysis research is experiencing rapid developments and making significant progress. Therefore, it is both necessary and challenging to comprehensively review the relevant studies. Although there are already some valuable and impressive surveys on enterprise risk analysis from the perspective of Finance and Management, these surveys introduce approaches in a relatively isolated way and lack recent advances in enterprise financial risk analysis. In contrast, this paper attempts to provide a systematic literature survey of enterprise risk analysis approaches from the perspective of Big Data and large language models. Specifically, this survey connects and systematizes existing research on enterprise financial risk, offering a holistic synthesis of research methods and key insights. We first introduce the problem formulation of enterprise financial risk in terms of risk types, granularity, intelligence levels, and evaluation metrics, and summarize representative studies accordingly. We then compare the analytical methods used to model enterprise financial risk and highlight the most influential research contributions. Finally, we identify the limitations of current research and propose five promising directions for future investigation.

Open set domain adaptation (OSDA) faces two critical challenges: the emergence of unknown classes in the target domain and changes in observed distributions across domains. Although numerous studies have proposed advanced algorithms, recent experimental results demonstrate that the classical empirical risk minimization (ERM) approach still delivers state-of-the-art performance. However, few theories can effectively explain this disputed phenomenon. To address the theoretical gap, we focus on constructing a causal theoretical framework for OSDA. We formulate the novel concepts of the fully informative causal invariance model (FICIM) and the partially informative causal invariance model (PICIM). Subsequently, we derive an OSDA theoretical bound to prove that the ERM performs well when the source domain follows FICIM, while it performs poorly when the source domain follows PICIM. The different results may be attributed to the varying amounts of available information when bounding the target domain's stable expected risk. Finally, across different datasets, we conduct extensive experiments on the FICIM and PICIM source domains to validate the effectiveness of our theoretical results. Moreover, our findings can also support the training and fine-tuning of large language models (LLMs).

This study addresses a critical gap in the literature regarding the lack of dynamic and causality sensitive decision support frameworks for carbon footprint based sustainable energy investments. Existing studies largely rely on static expert based or conventional fuzzy decision-making models, which are limited in capturing expert heterogeneity, learning dynamics, and complex interdependencies among strategic factors. To overcome these limitations, this paper proposes an integrated framework that combines Q learning with molecular fuzzy cognitive maps and a fuzzy molecular ranking approach. Q learning is employed to balance expert evaluation matrices by dynamically adjusting the judgments of less experienced decision makers based on reinforcement learning principles, thereby improving the consistency and reliability of expert inputs. The proposed model is empirically validated through a real-world expert-based case study involving three decision makers with heterogeneous experience levels and five alternative sustainable energy investment strategies evaluated under economic, environmental, social, and technical criteria. Molecular fuzzy cognitive maps enable the modeling of nonlinear causal relationships and uncertainty through geometry-based normalization, enhancing the robustness and adaptability of the weighting process across different learning rates and structural assumptions. Compared to conventional hybrid fuzzy MCDM models, the proposed framework demonstrates higher result stability and methodological flexibility while preserving interpretability. The results confirm the practical applicability of the model and provide actionable insights for policymakers and investors, identifying early-stage renewable energy startups as the most impactful strategy for reducing carbon footprints.

Revealing the underlying causal mechanisms in the real world is crucial for scientific and technological progress. Despite notable advances in recent decades, the lack of high-quality data and the reliance of traditional causal discovery algorithms (TCDA) on the assumption of no latent confounders, as well as their tendency to overlook the precise semantics of latent variables, have long been major obstacles to the broader application of causal discovery. To address this issue, we propose a novel causal modeling framework, TLVD, which integrates the metadata-based reasoning capabilities of large language models (LLMs) with the data-driven modeling capabilities of TCDA for inferring latent variables and their semantics. Specifically, we first employ a data-driven approach to construct a causal graph that incorporates latent variables. Then, we employ multi-LLM collaboration for latent variable inference, modeling this process as a game with incomplete information and seeking its Bayesian Nash Equilibrium (BNE) to infer the possible specific latent variables. Finally, to validate the inferred latent variables across multiple real-world web-based data sources, we leverage LLMs for evidence exploration to ensure traceability. We comprehensively evaluate TLVD on three de-identified real patient datasets provided by a hospital and two benchmark datasets. Extensive experimental results confirm the effectiveness and reliability of TLVD, with average improvements of 32.67% in Acc, 62.21% in CAcc, and 26.72% in ECit across the five datasets.

In group decision-making (GDM), the consensus-reaching process (CRP) is essential for aligning the diverse opinions of decision-makers (DMs) to achieve collective agreements. However, the process often faces obstacles due to the uncertainty of DMs in terms of unit cost and willingness to adjust opinions. To this end, this study constructs a new GDM framework by introducing reinforcement learning (RL) to the CRP. In this framework, we design a unit cost learning algorithm based on RL. The algorithm introduces an action space based on linguistic expressions, and therefore exhibits strong interpretability. On this basis, a weight reward-penalty mechanism based on asymmetric Nash bargaining is further proposed. The mechanism takes marginal and adjustment contributions as objective criteria, which provides a reasonable basis for improving consensus outcomes and managing non-cooperative behaviors. The proposed model incorporates both interactive and automatic strategies: the former is able to accurately capture individuals' willingness to cooperate with the help of RL, and the latter relies on optimization models to effectively reduce the time and cost spent on reaching consensus. Finally, we provide an example to illustrate the proposed approach and experimentally verify its feasibility and the potential of the RL framework.

This study aims to identify and prioritize optimal investment strategies for transforming remanufactured electronic waste into green building materials. Due to the complexity and uncertainty involved in such decisions, a novel hybrid decision-making model is proposed by integrating a Q-learning algorithm, a molecular fuzzy Bayesian network (BANEW), and a molecular ranking method (MORAN). The Q-learning algorithm is employed to objectively determine expert weights based on their experience and contribution levels, while molecular fuzzy sets are used to effectively handle uncertainty. The BANEW approach enables the identification of interdependencies among criteria, and the MORAN method is applied to rank the strategic alternatives. The findings indicate that efficiency is the most critical criterion with a weight of 0.313, followed by durability, environmental safety, and recyclability. Among the evaluated strategies, creating hybrid materials by blending e-waste is identified as the most optimal alternative with the highest aggregated value (0.464), followed by smart systems and standardization strategies. The results are validated through comparative analysis using the MAIRCA method and sensitivity analysis, both of which confirm the robustness and consistency of the proposed model. The main contribution of this study lies in integrating reinforcement learning with molecular fuzzy decision-making techniques to improve the accuracy of expert evaluation and reduce uncertainty in complex sustainability-related investment decisions. The proposed framework provides practical insights for policymakers and investors aiming to enhance sustainability in e-waste management and green building applications.