Agent Framework Research Articles

Existence of Optimal Stationary Singular Controls and Mean Field Game Equilibria

In this paper, we examine the stationary relaxed singular control problem within a multidimensional framework for a single agent as well as its mean field game equivalent. We demonstrate that optimal relaxed controls exist for two problem classes: one driven by queueing control and the other by harvesting models. These relaxed controls are defined by random measures across the state and control spaces with the state process described as a solution to the associated martingale problem. By leveraging findings from Kurtz and Stockbridge (2001), we establish the equivalence between the martingale problem and the stationary forward equation. This allows us to reformulate the relaxed control problem into a linear programming problem within the measure space. We prove the sequential compactness of these measures, thereby confirming the feasibility of achieving an optimal solution. Subsequently, our focus shifts to mean field games. Drawing on insights from the single-agent problem and employing the Kakutani–Glicksberg–Fan fixed point theorem, we derive the existence of a mean field game equilibria.

Conversational health agents: a personalized large language model-powered agent framework

ObjectiveConversational Health Agents (CHAs) are interactive systems providing healthcare services, such as assistance and diagnosis. Current CHAs, especially those utilizing Large Language Models (LLMs), primarily focus on conversation aspects. However, they offer limited agent capabilities, specifically needing more multistep problem-solving, personalized conversations, and multimodal data analysis. We aim to overcome these limitations.Materials and methodsWe propose openCHA, an open-source LLM-powered framework, designed to enable the development of conversational agents. OpenCHA offers a foundational and structured architecture and codebase, enabling researchers and developers to build and customize their CHA based on the specifics of their intended application. The framework leverages knowledge acquisition, problem-solving capabilities, multilingual, and multimodal conversations, and allows interaction with various AI platforms. We have released the framework as open source for the community on GitHub (https://github.com/Institute4FutureHealth/CHA and https://opencha.com).ResultsWe demonstrated the openCHA’s capability to develop CHAs across multiple health domains using 2 demos and 5 use cases. In diabetic patient management, developed CHA achieved a 92.1% accuracy rate, surpassing GPT4’s 51.8%. In food recommendations, developed CHA outperformed GPT4. The developed CHA excelled as an evaluator for mental health chatbots, recording the lowest Mean Absolute Error at 0.31, compared to competitors like GPT, Misteral, Gemini, and Claude. Additionally, the empathy enabled CHA identified emotional states with 89% accuracy, and in physiological data analysis of heart rate from Photoplethysmography (PPG) signals, the developed CHA achieved an mean absolute error of 2.83, far lower than GPT-4o’s 8.93.DiscussionThe openCHA framework enhances CHAs by enabling features such as explainability, personalization, and reliability through its integration with LLMs and external data sources. The developed CHAs face challenges like latency, token limits, and scalability. Future efforts will focus on improving planning robustness, enhancing accuracy and evaluation methods, and resolving user query ambiguity to further refine the framework’s effectiveness.ConclusionThe diverse demos and use cases of openCHA demonstrate the framework’s capacity to empower the development of a wide range of CHAs for various healthcare tasks.

Artificial Intelligence Agent Frameworks in Financial Stability: Innovations, Challenges, Applications

Artificial Intelligence (AI) agents are revolutionizing industries by enabling autonomous decision-making, task execution, multi-agent collaboration. This paper provides a comprehensive review of AI agent frameworks, focusing on their architectures, applications, challenges in financial services. We conduct a comparative analysis of leading frameworks, including LangGraph, CrewAI, AutoGen, evaluating their strengths, limitations, suitability for complex financial tasks such as trading, risk assessment, investment analysis. The integration of AI agents in financial markets presents both opportunities challenges, particularly in terms of regulatory compliance, ethical considerations, model robustness. We examine agentic AI design patterns, multi-agent systems, the deployment of AI agents advancing the proposal to use them for fraud detection risk management. By synthesizing insights from academic research industry practices, this review identifies key trends future directions in AI agent development. This work contributes to the growing discourse on AI-driven automation by outlining technical considerations open challenges in deploying AI agents at scale. We highlight the need for enhanced transparency, interpretability, security in AI-driven Agentic systems. Our findings provide valuable insights for researchers practitioners seeking to harness AI agents for more efficient intelligent decision-making.

Optimizing Recruitment with Machine Learning: A Novel Intelligent Agent Framework for HRM

Purpose: The study proposes a framework for integrating intelligent agents (IA) into human resource management (HRM) to improve recruitment, screening, training, and decision-making. It utilizes machine learning and pattern recognition to enhance candidate search accuracy and efficiency, addressing the limitations of traditional Boolean methods.Study design/methodology/approach: The study develops an Intelligent Agent AI (IAI) system for recruitment, using reinforcement learning and Naïve Bayes to optimize decision-making. It compares the IAI system to traditional methods like RecruitEm and Merlin, evaluating performance in accuracy, time efficiency, and resource management.Sample and data: The study customizes the IAI system, integrates historical candidate data, and conducts pilot testing in real-world HR settings. Job seeker data is used to train and test the system’s performance.Results: The IAI system significantly outperforms traditional methods in accuracy, time efficiency, and resource management. It also addresses issues like algorithmic bias and Boolean search limitations, improving candidate-job alignment and recruitment efficiency.Originality/value: The study provides a novel AI-powered solution to HR workflows, improving scalability, efficiency, and adaptability. It helps overcome traditional recruitment bias and enhances decision-making through data-driven insights.Research limitations/implications: Challenges include scalability, data bias, and integration issues. Future research should focus on addressing these limitations, refining AI processes, and optimizing performance in diverse HR contexts.

Optimizing Patient-AI Dialogue: Integrating Empathic Technology.

Patient empowerment is a growing focus in medical informatics, and artificial intelligence (AI) offers new opportunities for personalized, patient-centered support. This vision paper introduces a theoretical framework for an advanced medical conversational agent. We propose a conceptual model called Holistic Empathic AI (HEAI), implemented as a chatbot (PAI-bot). The system architecture integrates cognitive principles and includes modules for natural language dialogue, user state detection, and adaptive response generation. A key element is a multidimensional well-being score combining physical, emotional, cognitive, and social factors. The model structure and data flow support dynamic adjustment of interaction based on real-time analysis. Although not yet implemented clinically, this work establishes the conceptual foundation for future development of empathic AI in healthcare. The HEAI model aims to enhance the patient-AI dialogue and enable more human-like support.

A Framework of Hybrid System Dynamics and Agent Based Model With Cooperative Game Theory for Sustainable Coffee Supply Chain

The global coffee supply chain is a complex network involving diverse stakeholders such as farmers, traders, exporters, and consumers, each with unique incentives and constraints. This study introduces a conceptual framework that integrates System Dynamics (SD), Agent-Based Modelling (ABM), and Cooperative Game Theory (CGT) to address challenges in profit allocation, coalition stability, and sustainability. SD provides macro-level insights into global trends such as price fluctuations and production dynamics, while ABM models individual decision-making processes. CGT complements these methods by facilitating fair payoff distribution and stable coalition formation. The framework is structured into problem identification, model development and mapping, and interaction mode selection, offering a comprehensive approach to understanding material, information, and decision flows. Using illustrative scenarios, the study demonstrates the framework’s potential to analyses trade-offs and long-term impacts on supply chain stability. Its practical implications could support policymakers and industry leaders in designing fair profit-sharing mechanisms, promoting stable cooperation among stakeholders, and enhancing the overall sustainability of coffee and other agri-food supply chains. Thus, the framework highlights its applicability as a conceptual tool for supporting decision-making and sustainability in coffee supply chains and beyond.

AgCV: An Agentic framework for automating computer vision application

AgCV: An Agentic framework for automating computer vision application

Developing a multi-layer agent framework to enhance AI-generated educational questions for cybersecurity

Abstract This study evaluates the quality of questions generated by large language models, such as ChatGPT, in the context of cybersecurity education, using the CompTIA Security + certification as a case study. By analyzing 360 human-authored questions across 17 chapters and comparing them with AI-generated counterparts, the study employs Bloom’s Taxonomy to assess the cognitive levels achieved. The results reveal that AI-generated questions are predominantly limited to lower-order cognitive tasks, such as remembering and understanding, with significant gaps in addressing higher-order cognitive tasks, including applying, analyzing, evaluating, and creating. To address these limitations, a multi-layer agent framework was developed and implemented on a website. This framework integrates the ChatGPT API and processes the generated questions through multiple stages of evaluation, aligning them with Bloom’s Taxonomy and enhancing their quality. The framework includes scenario-based refinements, domain-specific fine-tuning, and a feedback mechanism to iteratively improve the cognitive depth of the questions. The agent systematically aligns AI-generated content with higher Bloom’s levels, making the questions more robust and applicable to real-world contexts. Evaluation of the framework demonstrates notable advancements in the quality of AI-generated questions, achieving closer alignment with human-authored content in terms of cognitive complexity, scenario depth, and relevance. The agent achieved substantial improvements in generating higher-order cognitive tasks, addressing the limitations of baseline AI performance. This study provides a scalable foundation for leveraging large language models in high-stakes educational assessments, with implications for adaptive learning and enhanced question design across diverse domains.

Taming Unleashed Large Language Models With Blockchain for Massive Personalized Reliable Healthcare.

The digital health field's pursuit of massive, personalized healthcare continuously faces constraints from doctors' resources and capacity limitations. Recently, the emergence of large language models (LLMs), with their remarkable comprehension and processing abilities, has revolutionized digital health and enhanced massive, personalized healthcare. Although these LLMs have achieved significant advancements, they have also introduced inevitable hallucinations, which impact patient safety when used in massive applications. To address these challenges, this study proposes a digital hospital for a massive, personalized, reliable healthcare service named the Chat Chain-Brain-based Doctor (CHATCBD). In addition, this study transforms the LLM-based diagnostic process into a digital hospital architecture, designs a controllable AI agents framework, and develops a self-audit mechanism to enhance their reliability. The proposed CHATCBD uses blockchain technology to decentralize external regulation of the LLMs' personalized diagnoses. It introduces a blockchain-based personalized routing management mechanism to improve patient-centered decision-making and designs a blockchain-based audit framework based on a proposed mathematical model that ensures both the professionalism and honesty of audits, serving as a safety net for addressing LLM hallucinations. The results of extensive experiments conducted on 13 datasets from multiple perspectives demonstrate that the proposed CHATCBD system can significantly enhance the capabilities of LLMs in personalized healthcare.

Agentic AI with retrieval-augmented generation for automated compliance assistance in finance

Maintaining compliance with complex Know Your Customer (KYC) and Anti-Money Laundering (AML) regulations is a resource-intensive challenge for financial institutions. This paper presents an agentic AI approach that leverages Retrieval-Augmented Generation (RAG) to automate and enhance compliance research and decision-making. We define the inefficiencies in current U.S. KYC/AML compliance workflows – including lengthy onboarding times and costly manual processes – as motivation for a more dynamic solution. We then introduce an autonomous agent framework, implemented with LangChain, that integrates a RAG pipeline to perform contextual reasoning over regulatory knowledge bases. The technical architecture is detailed with an emphasis on the agent’s planning and tool use capabilities, and the RAG components for knowledge base construction (using U.S. regulations such as FinCEN guidance, Code of Federal Regulations (CFR) provisions, and OFAC sanctions data), transformer-based embedding and indexing, vector retrieval, and LLM-driven answer generation. We demonstrate how this agent can handle compliance queries (e.g., customer due diligence requirements and detection of transaction structuring) in a simulated proof-of-concept. We discuss key advantages of this approach over traditional rule-based or static NLP systems – notably greater adaptability to changing regulations, improved traceability via source citations, and higher precision in complex scenario handling. Finally, we address ethical considerations (hallucination risk, ensuring regulatory accuracy, and model governance) and explore practical applications such as automated audit support, compliance report drafting, and future directions including real-time monitoring and multimodal compliance agents.

From DataOps to AIOps: How autonomous agents are revolutionizing data engineering

This comprehensive article examines the paradigm shift from traditional DataOps to AI-powered DataOps (AIOps), highlighting how autonomous agents are fundamentally transforming data engineering practices. The evolution represents not merely a technological upgrade but a complete reimagining of data pipeline management—moving from human-centered operations to self-learning, autonomous systems. The article explores the core pillars of AIOps: automated observability that contextually understands metrics beyond simple collection, predictive issue resolution that anticipates and prevents problems before they impact operations, and AI-driven metadata management that creates comprehensive knowledge graphs. It introduces the agentic framework comprising horizontal agents (resource optimization, performance monitoring, cost management, and security) and vertical agents (data quality, governance, domain-specific, and lineage tracking) that collaborate to create a truly intelligent ecosystem. The article further examines self-healing pipelines and emerging trends, including LLM-powered conversational interfaces, self-optimizing pipelines, and generative AI for documentation, while providing a phased implementation roadmap for organizations beginning their AIOps journey.

AI as Collaborative Partner: Rethinking Human-AI Teaming for the Real World

Much work in human-AI teaming today involves collaboration under fairly constrained settings. Humans supervise AI agents, who are relegated to following orders. The division of tasks is relatively superficial, with independent actions executed in parallel or with simple, linear dependencies between them. Communication is rigid and turn-based, with both parties speaking in complete, unambiguous sentences. However, collaboration in dynamic, real-world environments is rarely this straightforward. Team members adopt different roles as they continually adapt to an evolving situation, using efficient, timely communication to coordinate their actions. In this paper, we explore the technology requirements for AI agents to achieve this kind of collaboration and introduce COLLEAGUE, an agent framework designed to support adaptive, mixed-initiative interaction in dynamic domains.

A Comprehensive Review of Gen AI Agents: Applications and Frameworks in Finance, Investments and Risk Domains

This paper surveys the landscape of AI agent frameworks, highlights their core features and differences, and explores their applications in financial services. We synthesize insights from recent industry reports, academic research, and technical blog posts, focusing on frameworks such as CrewAI, LangGraph, LlamaIndex, and others. We also discuss the challenges and opportunities of deploying agentic AI in production environments, with an emphasis on financial trading, investment analysis, and decision support. We analyze the rapidly evolving landscape of agentic AI systems, focusing on their architecture, capabilities, and practical implementations in banking, trading, and risk management. The study examines prominent frameworks including LangGraph for stateful agent orchestration, CrewAI for collaborative multi-agent workflows, and AutoGen for conversational agent systems, alongside industry platforms like IBM watsonx and NVIDIA NIM. The study examines both technical frameworks (LangGraph, CrewAI, AutoGen, etc.) and practical implementations in financial institutions. We highlight productivity gains (up to 80% time reduction in data tasks), risk management improvements, and workforce transformation challenges. The paper concludes with recommendations for financial institutions adopting agentic AI solutions. Our analysis reveals three key findings: (1) specialized agent frameworks achieve 50-80% productivity gains in financial data tasks compared to traditional approaches, (2) multi-agent systems demonstrate particular promise in complex domains like algorithmic trading and fraud detection, and (3) successful deployment requires addressing critical challenges in workforce upskilling, risk alignment, and regulatory compliance. The paper provides a theoretical foundation for agentic AI in finance, introducing formal models for agent design patterns, multimodal fusion, and market microfoundations. We further present a summary of several evaluation frameworks for assessing agent performance across financial use cases, including portfolio optimization and AML compliance. The study concludes with recommendations for financial institutions adopting agentic AI, emphasizing the need for standardized architectures, robust testing protocols, and hybrid human-AI workflows.

Event-Driven Edge Agent Framework for Distributed Control in Distribution Networks

With the large-scale integration of heterogeneous energy resources and the increasing demand for flexible control, centralized control is facing challenges in terms of operational efficiency and system responsiveness when handling high-precision regulation tasks. To address this issue, this paper proposes an event-driven edge agent framework for distributed control in power distribution networks. First, based on the diverse requirements of distributed control in distribution networks, an edge agent architecture is constructed with modular components such as configuration management at its core. Second, considering the hybrid system characteristics of distribution networks, a control configuration technique based on activity-on-edge is designed, which decouples and discretizes continuous control processes through event-driven mechanisms. Furthermore, an edge-oriented automatic differentiation solver and a lightweight web application framework are developed to address the challenges of real-time optimization under resource-constrained environments. Finally, a semi-physical simulation is conducted using station-level economic dispatch as a case study to verify the effectiveness of the proposed technology. The results demonstrate that, compared to centralized control, the designed distributed agent maintains optimization accuracy while reducing event-triggering frequency by 40% and improving communication response speed by 70%, showing strong performance in operational efficiency at the edge.

Research on Mobile Agent Path Planning Based on Deep Reinforcement Learning

For mobile agent path planning, traditional path planning algorithms frequently induce abrupt variations in path curvature and steering angles, increasing the risk of lateral tire slippage and undermining operational safety. Concurrently, conventional reinforcement learning methods struggle to converge rapidly, leading to an insufficient efficiency in planning to meet the demand for energy economy. This study proposes LSTM Bézier–Double Deep Q-Network (LB-DDQN), an advanced path-planning framework for mobile agents based on deep reinforcement learning. The architecture first enables mapless navigation through a DDQN foundation, subsequently integrates long short-term memory (LSTM) networks for the fusion of environmental features and preservation of training information, and ultimately enhances the path’s quality through redundant node elimination via an obstacle–path relationship analysis, combined with Bézier curve-based trajectory smoothing. A sensor-driven three-dimensional simulation environment featuring static obstacles was constructed using the ROS and Gazebo platforms, where LiDAR-equipped mobile agent models were trained for real-time environmental perception and strategy optimization prior to deployment on experimental vehicles. The simulation and physical implementation results reveal that LB-DDQN achieves effective collision avoidance, while demonstrating marked enhancements in critical metrics: the path’s smoothness, energy efficiency, and motion stability exhibit average improvements exceeding 50%. The framework further maintains superior safety standards and operational efficiency across diverse scenarios.

SWARM: Reimagining scientific workflow management systems in a distributed world

Modern scientific workflows process massive amounts of data from diverse instruments and sensors, leveraging geographically distributed, heterogeneous compute and storage resources—from leadership-class systems to edge devices—connected by high-performance networks. The diversity of resources introduces challenges in harnessing their full potential, with resilience issues arising across applications, system software, networks, storage, and hardware. Today, workflow management systems (WMS) coordinate the execution of computation and data management tasks across target resources. However, WMS’s centralized nature makes them vulnerable to faults and scalability issues that may result in failures of entire computational campaigns. This paper introduces a novel agentic framework for workflow management, fully distributing and decentralizing the WMS functions and modeling them as swarm intelligence agents infused with advanced artificial intelligence solutions and traditional distributed computing algorithms that can make coordinated decisions in the presence of failures of the underlying cyberinfrastructure.

An Intention-Aware Agent Framework for Multi-Agent Decentralized Partially Observable Environments

In the real world, humans often collaborate with others without direct communication. To do this successfully, they have to infer their intentions and choose actions that complement the predicted actions of their collaborators to perform the task efficiently. Since the peer’s state and action are generally not directly observable, these are usually estimated based on environmental change and then used to predict the intention. While humans can achieve this easily, this form of collaboration is difficult for artificial intelligent agents operating in partially observable environments, leading to agent architectures that do not attempt to explicitly infer other agents’ intentions but rather rely on additional knowledge or reactive collaboration, relying on the steadystate character of other agents.In this paper, we propose an agent model that explicitly defines and utilizes estimates of other agents’ intentions to yield more effective collaboration in decentralized partially observable domains, where each agent’s knowledge of and current belief state in the environment can be different. The resulting agents explicitly estimate other agents’ intentions from their observationsand utilize these estimates in a Reinforcement Learning process on a modified Dec-POMDP model to learn collaborative strategies. Initial experiments in a simple, partially observable collaborative manipulation domain show the ability of these intention-aware agents to learn optimal hierarchical strategies faster and more stably than equivalent agents without intention awareness.

Customized Application of Intelligent Agents in Industrial Robotic Smart Production Lines

Intelligent agent technology, as a core driver of industrial intelligent transformation, enables customized, adaptive, and highly efficient collaboration in industrial robotic production lines by integrating knowledge bases, workflow engines, and extensible plugin systems. This paper begins with the technical framework of intelligent agents, analyzes their customized application scenarios in industrial robotic smart production lines-including knowledge-driven task scheduling, dynamic workflow optimization, and multi-device collaborative control-and explores how intelligent agents enhance production line flexibility and intelligence through hardware interface design and software functional modules.

Evaluation framework for conversational AI agents in pharmacy education: A scoping review of key characteristics and outcome measures.

Evaluation framework for conversational AI agents in pharmacy education: A scoping review of key characteristics and outcome measures.

Heterogeneous Agent Models

Heterogeneous agent models have become central to modern macroeconomic research, often replacing the representative agent framework. However, what is core for these frameworks is the use of microfoundations that involve optimizing behavior. The strength of heterogeneous agent models lies in their ability to address questions where heterogeneity is essential--for instance, examining renters versus homeowners, the old versus the young, or the rich versus the poor. While advances in computational power have allowed these models to replicate observed wealth and income distributions effectively, much remains to be understood about the bottom of the income and wealth distributions.