Does continued scaling of large language models (LLMs) yield diminishing returns? Real-world value often stems from the length of task an agent can complete. We start this work by observing the simple but counterintuitive fact that marginal gains in single-step accuracy can compound into exponential improvements in the length of a task a model can successfully complete. Then, we argue that failures of LLMs when simple tasks are made longer arise from mistakes in execution, rather than an inability to reason. We propose isolating execution capability, by explicitly providing the knowledge and plan needed to solve a long-horizon task. We find that larger models can correctly execute significantly more turns even when small models have 100\% single-turn accuracy. We observe that the per-step accuracy of models degrades as the number of steps increases. This is not just due to long-context limitations -- curiously, we observe a self-conditioning effect -- models become more likely to make mistakes when the context contains their errors from prior turns. Self-conditioning does not reduce by just scaling the model size. In contrast, recent thinking models do not self-condition, and can also execute much longer tasks in a single turn. We conclude by benchmarking frontier thinking models on the length of task they can execute in a single turn. Overall, by focusing on the ability to execute, we hope to reconcile debates on how LLMs can solve complex reasoning problems yet fail at simple tasks when made longer, and highlight the massive benefits of scaling model size and sequential test-time compute for long-horizon tasks.
The advancement of large language models (LLMs) has enabled the construction of multi-agent systems to solve complex tasks by dividing responsibilities among specialized agents, such as a planning agent for subgoal generation and a grounding agent for executing tool-use actions. Most existing methods typically fine-tune these agents independently, leading to capability gaps among them with poor coordination. To address this, we propose MOAT, a Multi-Agent Joint Alignment Tuning framework that improves agents collaboration through iterative alignment. MOAT alternates between two key stages: (1) Planning Agent Alignment, which optimizes the planning agent to generate subgoal sequences that better guide the grounding agent; and (2) Grounding Agent Improving, which fine-tunes the grounding agent using diverse subgoal-action pairs generated by the agent itself to enhance its generalization capablity. Theoretical analysis proves that MOAT ensures a non-decreasing and progressively convergent training process. Experiments across six benchmarks demonstrate that MOAT outperforms state-of-the-art baselines, achieving average improvements of 3.1% on held-in tasks and 4.4% on held-out tasks.
Serverless computing provides dynamic scalability, but its infrastructure overhead becomes a bottleneck for emerging workloads such as LLM agents, which exhibit unpredictable invocation patterns and variable resource demands. Our analysis shows that for these agents, the cost of running on serverless platforms can reach up to 70% of the cost of LLM API calls. This finding motivates the need for a more efficient, high-density serverless platform. We present TrEnv, a co-designed serverless platform that supports both container- and VM-based environments, optimized for the unique demands of LLM agents. TrEnv reduces startup latency and memory usage through repurposable sandboxes and memory templates, which enable fast reuse and restoration of execution environments. To further reduce overhead in VM-based agent workloads, TrEnv leverages browser sharing and a page cache bypassing mechanism. Evaluations show that TrEnv reduces P99 latency by up to 7X and memory usage by 48% in container-based settings, and achieves up to 58% lower P99 latency and 61% memory savings for VM-based agents compared to state-of-the-art systems like E2B.
LLMs now form the backbone of AI agents for a diverse array of applications, including tool use, command-line agents, and web or computer use agents. These agentic LLM inference tasks are fundamentally different from chatbot-focused inference -- they often have much larger context lengths to capture complex, prolonged inputs, such as entire webpage DOMs or complicated tool call trajectories. This, in turn, generates significant off-chip memory traffic for the underlying hardware at the inference stage and causes the workload to be constrained by two memory walls, namely the bandwidth and capacity memory walls, preventing the on-chip compute units from achieving high utilization. In this paper, we introduce PLENA, a hardware-software co-designed system that applies three core optimization pathways to tackle these challenges. PLENA includes an efficient hardware implementation of compute and memory units supporting an asymmetric quantization scheme. PLENA also features a novel flattened systolic array architecture that has native support for FlashAttention to tackle these memory walls in the scenario of inference serving for long-context LLMs. Additionally, PLENA is developed with a complete stack, including a custom ISA, a compiler, a cycle-emulated simulator, and an automated design space exploration flow. The simulated results show that PLENA achieves up to 8.5x higher utilization than existing accelerators, and delivers 2.24x higher throughput than the A100 GPU and 3.85x higher throughput than the TPU v6e, under the same multiplier count and memory settings. The full PLENA system will also be open-sourced.
Large Language Models (LLMs) are increasingly deployed in enterprise applications, yet their reliability remains limited by hallucinations, i.e., confident but factually incorrect information. Existing detection approaches, such as SelfCheckGPT and MetaQA, primarily target standalone LLMs and do not address the unique challenges of Retrieval-Augmented Generation (RAG) systems, where responses must be consistent with retrieved evidence. We therefore present MetaRAG, a metamorphic testing framework for hallucination detection in Retrieval-Augmented Generation (RAG) systems. MetaRAG operates in a real-time, unsupervised, black-box setting, requiring neither ground-truth references nor access to model internals, making it suitable for proprietary and high-stakes domains. The framework proceeds in four stages: (1) decompose answers into atomic factoids, (2) generate controlled mutations of each factoid using synonym and antonym substitutions, (3) verify each variant against the retrieved context (synonyms are expected to be entailed and antonyms contradicted), and (4) aggregate penalties for inconsistencies into a response-level hallucination score. Crucially for identity-aware AI, MetaRAG localizes unsupported claims at the factoid span where they occur (e.g., pregnancy-specific precautions, LGBTQ+ refugee rights, or labor eligibility), allowing users to see flagged spans and enabling system designers to configure thresholds and guardrails for identity-sensitive queries. Experiments on a proprietary enterprise dataset illustrate the effectiveness of MetaRAG for detecting hallucinations and enabling trustworthy deployment of RAG-based conversational agents. We also outline a topic-based deployment design that translates MetaRAG's span-level scores into identity-aware safeguards; this design is discussed but not evaluated in our experiments.
Recent advances in large language models (LLMs) have enabled the emergence of general-purpose agents for automating end-to-end machine learning (ML) workflows, including data analysis, feature engineering, model training, and competition solving. However, existing benchmarks remain limited in task coverage, domain diversity, difficulty modeling, and evaluation rigor, failing to capture the full capabilities of such agents in realistic settings. We present TAM Bench, a diverse, realistic, and structured benchmark for evaluating LLM-based agents on end-to-end ML tasks. TAM Bench features three key innovations: (1) A browser automation and LLM-based task acquisition system that automatically collects and structures ML challenges from platforms such as Kaggle, AIcrowd, and Biendata, spanning multiple task types and data modalities (e.g., tabular, text, image, graph, audio); (2) A leaderboard-driven difficulty modeling mechanism that estimates task complexity using participant counts and score dispersion, enabling scalable and objective task calibration; (3) A multi-dimensional evaluation framework incorporating performance, format compliance, constraint adherence, and task generalization. Based on 150 curated AutoML tasks, we construct three benchmark subsets of different sizes -- Lite, Medium, and Full -- designed for varying evaluation scenarios. The Lite version, with 18 tasks and balanced coverage across modalities and difficulty levels, serves as a practical testbed for daily benchmarking and comparative studies.
Materials characterization is fundamental to acquiring materials information, revealing the processing-microstructure-property relationships that guide material design and optimization. While multimodal large language models (MLLMs) have recently shown promise in generative and predictive tasks within materials science, their capacity to understand real-world characterization imaging data remains underexplored. To bridge this gap, we present MatCha, the first benchmark for materials characterization image understanding, comprising 1,500 questions that demand expert-level domain expertise. MatCha encompasses four key stages of materials research comprising 21 distinct tasks, each designed to reflect authentic challenges faced by materials scientists. Our evaluation of state-of-the-art MLLMs on MatCha reveals a significant performance gap compared to human experts. These models exhibit degradation when addressing questions requiring higher-level expertise and sophisticated visual perception. Simple few-shot and chain-of-thought prompting struggle to alleviate these limitations. These findings highlight that existing MLLMs still exhibit limited adaptability to real-world materials characterization scenarios. We hope MatCha will facilitate future research in areas such as new material discovery and autonomous scientific agents. MatCha is available at https://github.com/FreedomIntelligence/MatCha.
With the rapid advancement of large language models (LLMs), Multi-agent Systems (MAS) have achieved significant progress in various application scenarios. However, substantial challenges remain in designing versatile, robust, and efficient platforms for agent deployment. To address these limitations, we propose \textbf{LightAgent}, a lightweight yet powerful agentic framework, effectively resolving the trade-off between flexibility and simplicity found in existing frameworks. LightAgent integrates core functionalities such as Memory (mem0), Tools, and Tree of Thought (ToT), while maintaining an extremely lightweight structure. As a fully open-source solution, it seamlessly integrates with mainstream chat platforms, enabling developers to easily build self-learning agents. We have released LightAgent at \href{https://github.com/wxai-space/LightAgent}{https://github.com/wxai-space/LightAgent}
In long-horizon tasks, recent agents based on Large Language Models (LLMs) face a significant challenge that sparse, outcome-based rewards make it difficult to assign credit to intermediate steps. Previous methods mainly focus on creating dense reward signals to guide learning, either through traditional reinforcement learning techniques like inverse reinforcement learning or by using Process Reward Models for step-by-step feedback. In this paper, we identify a fundamental problem in the learning dynamics of LLMs: the magnitude of policy gradients is inherently coupled with the entropy, which leads to inefficient small updates for confident correct actions and potentially destabilizes large updates for uncertain ones. To resolve this, we propose Entropy-Modulated Policy Gradients (EMPG), a framework that re-calibrates the learning signal based on step-wise uncertainty and the final task outcome. EMPG amplifies updates for confident correct actions, penalizes confident errors, and attenuates updates from uncertain steps to stabilize exploration. We further introduce a bonus term for future clarity that encourages agents to find more predictable solution paths. Through comprehensive experiments on three challenging agent tasks, WebShop, ALFWorld, and Deep Search, we demonstrate that EMPG achieves substantial performance gains and significantly outperforms strong policy gradient baselines. Project page is at https://empgseed-seed.github.io/
Large language models (LLMs) have shown great promise in automating data science workflows, but existing models still struggle with multi-step reasoning and tool use, which limits their effectiveness on complex data analysis tasks. To address this, we propose a scalable pipeline that extracts high-quality, tool-based data analysis tasks and their executable multi-step solutions from real-world Jupyter notebooks and associated data files. Using this pipeline, we introduce NbQA, a large-scale dataset of standardized task-solution pairs that reflect authentic tool-use patterns in practical data science scenarios. To further enhance multi-step reasoning, we present Jupiter, a framework that formulates data analysis as a search problem and applies Monte Carlo Tree Search (MCTS) to generate diverse solution trajectories for value model learning. During inference, Jupiter combines the value model and node visit counts to efficiently collect executable multi-step plans with minimal search steps. Experimental results show that Qwen2.5-7B and 14B-Instruct models on NbQA solve 77.82% and 86.38% of tasks on InfiAgent-DABench, respectively-matching or surpassing GPT-4o and advanced agent frameworks. Further evaluations demonstrate improved generalization and stronger tool-use reasoning across diverse multi-step reasoning tasks.
Question Answering over Tabular Data (Table QA) presents unique challenges due to the diverse structure, size, and data types of real-world tables. The SemEval 2025 Task 8 (DataBench) introduced a benchmark composed of large-scale, domain-diverse datasets to evaluate the ability of models to accurately answer structured queries. We propose a Natural Language to SQL (NL-to-SQL) approach leveraging large language models (LLMs) such as GPT-4o, GPT-4o-mini, and DeepSeek v2:16b to generate SQL queries dynamically. Our system follows a multi-stage pipeline involving example selection, SQL query generation, answer extraction, verification, and iterative refinement. Experiments demonstrate the effectiveness of our approach, achieving 70.5\% accuracy on DataBench QA and 71.6\% on DataBench Lite QA, significantly surpassing baseline scores of 26\% and 27\% respectively. This paper details our methodology, experimental results, and alternative approaches, providing insights into the strengths and limitations of LLM-driven Table QA.
Large language models (LLMs)-empowered autonomous agents are transforming both digital and physical environments by enabling adaptive, multi-agent collaboration. While these agents offer significant opportunities across domains such as finance, healthcare, and smart manufacturing, their unpredictable behaviors and heterogeneous capabilities pose substantial governance and accountability challenges. In this paper, we propose a blockchain-enabled layered architecture for regulatory agent collaboration, comprising an agent layer, a blockchain data layer, and a regulatory application layer. Within this framework, we design three key modules: (i) an agent behavior tracing and arbitration module for automated accountability, (ii) a dynamic reputation evaluation module for trust assessment in collaborative scenarios, and (iii) a malicious behavior forecasting module for early detection of adversarial activities. Our approach establishes a systematic foundation for trustworthy, resilient, and scalable regulatory mechanisms in large-scale agent ecosystems. Finally, we discuss the future research directions for blockchain-enabled regulatory frameworks in multi-agent systems.
Large Language Models (LLMs) are fast becoming indispensable tools for software developers, assisting or even partnering with them in crafting complex programs. The advantages are evident -- LLMs can significantly reduce development time, generate well-organized and comprehensible code, and occasionally suggest innovative ideas that developers might not conceive on their own. However, despite their strengths, LLMs will often introduce significant errors and present incorrect code with persuasive confidence, potentially misleading developers into accepting flawed solutions. In order to bring LLMs into the software development cycle in a more reliable manner, we propose a methodology for combining them with ``traditional'' software engineering techniques in a structured way, with the goal of streamlining the development process, reducing errors, and enabling users to verify crucial program properties with increased confidence. Specifically, we focus on the Scenario-Based Programming (SBP) paradigm -- an event-driven, scenario-based approach for software engineering -- to allow human developers to pour their expert knowledge into the LLM, as well as to inspect and verify its outputs. To evaluate our methodology, we conducted a significant case study, and used it to design and implement the Connect4 game. By combining LLMs and SBP we were able to create a highly-capable agent, which could defeat various strong existing agents. Further, in some cases, we were able to formally verify the correctness of our agent. Finally, our experience reveals interesting insights regarding the ease-of-use of our proposed approach. The full code of our case-study will be made publicly available with the final version of this paper.
Artificial Intelligence (AI) techniques play a pivotal role in optimizing wireless communication networks. However, traditional deep learning approaches often act as closed boxes, lacking the structured reasoning abilities needed to tackle complex, multi-step decision problems. This survey provides a comprehensive review and outlook of reasoning-enabled AI in wireless communication networks, with a focus on Large Language Models (LLMs) and other advanced reasoning paradigms. In particular, LLM-based agents can combine reasoning with long-term planning, memory, tool utilization, and autonomous cross-layer control to dynamically optimize network operations with minimal human intervention. We begin by outlining the evolution of intelligent wireless networking and the limitations of conventional AI methods. We then introduce emerging AI reasoning techniques. Furthermore, we establish a classification system applicable to wireless network tasks. We also present a layer-by-layer examination for AI reasoning, covering the physical, data link, network, transport, and application layers. For each part, we identify key challenges and illustrate how AI reasoning methods can improve AI-based wireless communication performance. Finally, we discuss key research directions for AI reasoning toward future wireless communication networks. By combining insights from both communications and AI, this survey aims to chart a path for integrating reasoning techniques into the next-generation wireless networks.
Coordination tasks traditionally performed by humans are increasingly being delegated to autonomous agents. As this pattern progresses, it becomes critical to evaluate not only these agents' performance but also the processes through which they negotiate in dynamic, multi-agent environments. Furthermore, different agents exhibit distinct advantages: traditional statistical agents, such as Bayesian models, may excel under well-specified conditions, whereas large language models (LLMs) can generalize across contexts. In this work, we compare humans (N = 216), LLMs (GPT-4o, Gemini 1.5 Pro), and Bayesian agents in a dynamic negotiation setting that enables direct, identical-condition comparisons across populations, capturing both outcomes and behavioral dynamics. Bayesian agents extract the highest surplus through aggressive optimization, at the cost of frequent trade rejections. Humans and LLMs can achieve similar overall surplus, but through distinct behaviors: LLMs favor conservative, concessionary trades with few rejections, while humans employ more strategic, risk-taking, and fairness-oriented behaviors. Thus, we find that performance parity -- a common benchmark in agent evaluation -- can conceal fundamental differences in process and alignment, which are critical for practical deployment in real-world coordination tasks.
Natural language descriptions of optimization or satisfaction problems are challenging to translate into correct MiniZinc models, as this process demands both logical reasoning and constraint programming expertise. We introduce a framework that addresses this challenge with an agentic approach: multiple specialized large language model (LLM) agents decompose the modeling task by global constraint type. Each agent is dedicated to detecting and generating code for a specific class of global constraint, while a final assembler agent integrates these constraint snippets into a complete MiniZinc model. By dividing the problem into smaller, well-defined sub-tasks, each LLM handles a simpler reasoning challenge, potentially reducing overall complexity. We conduct initial experiments with several LLMs and show better performance against baselines such as one-shot prompting and chain-of-thought prompting. Finally, we outline a comprehensive roadmap for future work, highlighting potential enhancements and directions for improvement.
Large Language Models (LLMs), when paired with prompt-based tasks, have significantly reduced data annotation costs and reliance on human annotators. However, evaluating the quality of their annotations remains challenging in dynamic, unsupervised environments where oracle feedback is scarce and conventional methods fail. To address this challenge, we propose a novel agentic annotation paradigm, where a student model collaborates with a noisy teacher (the LLM) to assess and refine annotation quality without relying on oracle feedback. The student model, acting as an unsupervised feedback mechanism, employs a user preference-based majority voting strategy to evaluate the consistency of the LLM outputs. To systematically measure the reliability of LLM-generated annotations, we introduce the Consistent and Inconsistent (CAI) Ratio, a novel unsupervised evaluation metric. The CAI Ratio not only quantifies the annotation quality of the noisy teacher under limited user preferences but also plays a critical role in model selection, enabling the identification of robust LLMs in dynamic, unsupervised environments. Applied to ten open-domain NLP datasets across four LLMs, the CAI Ratio demonstrates a strong positive correlation with LLM accuracy, establishing it as an essential tool for unsupervised evaluation and model selection in real-world settings.
Developing autonomous LLM agents capable of making a series of intelligent decisions to solve complex, real-world tasks is a fast-evolving frontier. Like human cognitive development, agents are expected to acquire knowledge and skills through exploration and interaction with the environment. Despite advances, the community still lacks a unified, interactive reinforcement learning (RL) framework that can effectively train such agents from scratch -- without relying on supervised fine-tuning (SFT) -- across diverse and realistic environments. To bridge this gap, we introduce AgentGym-RL, a new framework to train LLM agents for multi-turn interactive decision-making through RL. The framework features a modular and decoupled architecture, ensuring high flexibility and extensibility. It encompasses a wide variety of real-world scenarios, and supports mainstream RL algorithms. Furthermore, we propose ScalingInter-RL, a training approach designed for exploration-exploitation balance and stable RL optimization. In early stages, it emphasizes exploitation by restricting the number of interactions, and gradually shifts towards exploration with larger horizons to encourage diverse problem-solving strategies. In this way, the agent develops more diverse behaviors and is less prone to collapse under long horizons. We perform extensive experiments to validate the stability and effectiveness of both the AgentGym-RL framework and the ScalingInter-RL approach. Our agents match or surpass commercial models on 27 tasks across diverse environments. We offer key insights and will open-source the complete AgentGym-RL framework -- including code and datasets -- to empower the research community in developing the next generation of intelligent agents.
The efficiency of Bayesian optimization (BO) in chemistry is often hindered by sparse experimental data and complex reaction mechanisms. To overcome these limitations, we introduce ChemBOMAS, a new framework named LLM-Enhanced Multi-Agent System for accelerating BO in chemistry. ChemBOMAS's optimization process is enhanced by LLMs and synergistically employs two strategies: knowledge-driven coarse-grained optimization and data-driven fine-grained optimization. First, in the knowledge-driven coarse-grained optimization stage, LLMs intelligently decompose the vast search space by reasoning over existing chemical knowledge to identify promising candidate regions. Subsequently, in the data-driven fine-grained optimization stage, LLMs enhance the BO process within these candidate regions by generating pseudo-data points, thereby improving data utilization efficiency and accelerating convergence. Benchmark evaluations** further confirm that ChemBOMAS significantly enhances optimization effectiveness and efficiency compared to various BO algorithms. Importantly, the practical utility of ChemBOMAS was validated through wet-lab experiments conducted under pharmaceutical industry protocols, targeting conditional optimization for a previously unreported and challenging chemical reaction. In the wet experiment, ChemBOMAS achieved an optimal objective value of 96%. This was substantially higher than the 15% achieved by domain experts. This real-world success, together with strong performance on benchmark evaluations, highlights ChemBOMAS as a powerful tool to accelerate chemical discovery.
Creating large-scale verifiable training datasets for issue-resolving tasks is a critical yet notoriously difficult challenge. Existing methods on automating the Gym environment setup process for real-world issues suffer from low success rates and high overhead. Meanwhile, synthesizing new tasks within existing Gym environments leaves the vast pool of authentic, human-reported problems untapped. To maximize the utilization of existing Gym environments and also the rich data of issue-resolving history on GitHub, we introduce SWE-Mirror, a pipeline that distills a real-world issue's semantic essence, mirrors it into another repository with a configured Gym environment, and re-animates it as a verifiable issue-resolving task. SWE-Mirror reuses existing Gym environments along with the vast pool of issue-resolving history hosted on GitHub to construct a large-scale dataset of mirrored authentic and verifiable tasks. Applying SWE-Mirror to 40 repositories across 4 languages, we have curated a dataset with 60,671 issue-resolving tasks and demonstrated the value of our dataset by training and evaluating coding agents at various scale. Post-training experiments show that models trained with the dataset exhibit improvements in issue-resolving capabilities. Furthermore, by extending the dataset size to over 12,000 high-quality trajectories, we established a new state-of-the-art (SOTA) among Qwen2.5-Coder-Instruct based LLMs on the OpenHands agent framework, which increases the resolve rate on SWE-Bench-Verified by +21.8% for the 7B model and +46.0% for the 32B model and validates the effectiveness of our approach.
As Large Language Model (LLM) agents become increasingly capable of automating complex, multi-step tasks, the need for robust, secure, and predictable architectural patterns is paramount. This paper provides a comprehensive guide to the ``Plan-then-Execute'' (P-t-E) pattern, an agentic design that separates strategic planning from tactical execution. We explore the foundational principles of P-t-E, detailing its core components - the Planner and the Executor - and its architectural advantages in predictability, cost-efficiency, and reasoning quality over reactive patterns like ReAct (Reason + Act). A central focus is placed on the security implications of this design, particularly its inherent resilience to indirect prompt injection attacks by establishing control-flow integrity. We argue that while P-t-E provides a strong foundation, a defense-in-depth strategy is necessary, and we detail essential complementary controls such as the Principle of Least Privilege, task-scoped tool access, and sandboxed code execution. To make these principles actionable, this guide provides detailed implementation blueprints and working code references for three leading agentic frameworks: LangChain (via LangGraph), CrewAI, and AutoGen. Each framework's approach to implementing the P-t-E pattern is analyzed, highlighting unique features like LangGraph's stateful graphs for re-planning, CrewAI's declarative tool scoping for security, and AutoGen's built-in Docker sandboxing. Finally, we discuss advanced patterns, including dynamic re-planning loops, parallel execution with Directed Acyclic Graphs (DAGs), and the critical role of Human-in-the-Loop (HITL) verification, to offer a complete strategic blueprint for architects, developers, and security engineers aiming to build production-grade, resilient, and trustworthy LLM agents.
Specialized machine learning models, regardless of architecture and training, are susceptible to failures in deployment. With their increasing use in high risk situations, the ability to audit these models by determining their operational design domain (ODD) is crucial in ensuring safety and compliance. However, given the high-dimensional input spaces, this process often requires significant human resources and domain expertise. To alleviate this, we introduce \coolname, an LLM-Agent centric framework for automated generation of semantically relevant test cases to search for failure modes in specialized black-box models. By leveraging LLM-Agents as tool orchestrators, we aim to fit a uncertainty-aware failure distribution model on a learned text-embedding manifold by projecting the high-dimension input space to low-dimension text-embedding latent space. The LLM-Agent is tasked with iteratively building the failure landscape by leveraging tools for generating test-cases to probe the model-under-test (MUT) and recording the response. The agent also guides the search using tools to probe uncertainty estimate on the low dimensional manifold. We demonstrate this process in a simple case using models trained with missing digits on the MNIST dataset and in the real world setting of vision-based intruder detection for aerial vehicles.
The substantial data volumes encountered in modern particle physics and other domains of fundamental physics research allow (and require) the use of increasingly complex data analysis tools and workflows. While the use of machine learning (ML) tools for data analysis has recently proliferated, these tools are typically special-purpose algorithms that rely, for example, on encoded physics knowledge to reach optimal performance. In this work, we investigate a new and orthogonal direction: Using recent progress in large language models (LLMs) to create a team of agents -- instances of LLMs with specific subtasks -- that jointly solve data analysis-based research problems in a way similar to how a human researcher might: by creating code to operate standard tools and libraries (including ML systems) and by building on results of previous iterations. If successful, such agent-based systems could be deployed to automate routine analysis components to counteract the increasing complexity of modern tool chains. To investigate the capabilities of current-generation commercial LLMs, we consider the task of anomaly detection via the publicly available and highly-studied LHC Olympics dataset. Several current models by OpenAI (GPT-4o, o4-mini, GPT-4.1, and GPT-5) are investigated and their stability tested. Overall, we observe the capacity of the agent-based system to solve this data analysis problem. The best agent-created solutions mirror the performance of human state-of-the-art results.
As humans delegate more tasks and decisions to artificial intelligence (AI), we risk losing control of our individual and collective futures. Relatively simple algorithmic systems already steer human decision-making, such as social media feed algorithms that lead people to unintentionally and absent-mindedly scroll through engagement-optimized content. In this paper, we develop the idea of human agency by integrating philosophical and scientific theories of agency with AI-assisted evaluation methods: using large language models (LLMs) to simulate and validate user queries and to evaluate AI responses. We develop HumanAgencyBench (HAB), a scalable and adaptive benchmark with six dimensions of human agency based on typical AI use cases. HAB measures the tendency of an AI assistant or agent to Ask Clarifying Questions, Avoid Value Manipulation, Correct Misinformation, Defer Important Decisions, Encourage Learning, and Maintain Social Boundaries. We find low-to-moderate agency support in contemporary LLM-based assistants and substantial variation across system developers and dimensions. For example, while Anthropic LLMs most support human agency overall, they are the least supportive LLMs in terms of Avoid Value Manipulation. Agency support does not appear to consistently result from increasing LLM capabilities or instruction-following behavior (e.g., RLHF), and we encourage a shift towards more robust safety and alignment targets.
Generating regulatorily compliant Suspicious Activity Report (SAR) remains a high-cost, low-scalability bottleneck in Anti-Money Laundering (AML) workflows. While large language models (LLMs) offer promising fluency, they suffer from factual hallucination, limited crime typology alignment, and poor explainability -- posing unacceptable risks in compliance-critical domains. This paper introduces Co-Investigator AI, an agentic framework optimized to produce Suspicious Activity Reports (SARs) significantly faster and with greater accuracy than traditional methods. Drawing inspiration from recent advances in autonomous agent architectures, such as the AI Co-Scientist, our approach integrates specialized agents for planning, crime type detection, external intelligence gathering, and compliance validation. The system features dynamic memory management, an AI-Privacy Guard layer for sensitive data handling, and a real-time validation agent employing the Agent-as-a-Judge paradigm to ensure continuous narrative quality assurance. Human investigators remain firmly in the loop, empowered to review and refine drafts in a collaborative workflow that blends AI efficiency with domain expertise. We demonstrate the versatility of Co-Investigator AI across a range of complex financial crime scenarios, highlighting its ability to streamline SAR drafting, align narratives with regulatory expectations, and enable compliance teams to focus on higher-order analytical work. This approach marks the beginning of a new era in compliance reporting -- bringing the transformative benefits of AI agents to the core of regulatory processes and paving the way for scalable, reliable, and transparent SAR generation.
Large Language Models (LLMs), with their strong understanding and reasoning capabilities, are increasingly being explored for tackling optimization problems, especially in synergy with evolutionary computation. Despite rapid progress, however, the field still lacks a unified synthesis and a systematic taxonomy. This survey addresses this gap by providing a comprehensive review of recent developments and organizing them within a structured framework. We classify existing research into two main stages: LLMs for optimization modeling and LLMs for optimization solving. The latter is further divided into three paradigms according to the role of LLMs in the optimization workflow: LLMs as stand-alone optimizers, low-level LLMs embedded within optimization algorithms, and high-level LLMs for algorithm selection and generation. For each category, we analyze representative methods, distill technical challenges, and examine their interplay with traditional approaches. We also review interdisciplinary applications spanning the natural sciences, engineering, and machine learning. By contrasting LLM-driven and conventional methods, we highlight key limitations and research gaps, and point toward future directions for developing self-evolving agentic ecosystems for optimization. An up-to-date collection of related literature is maintained at https://github.com/ishmael233/LLM4OPT.
LLMs have made substantial progress in task automation and natural language understanding.However,without expertise in GIS,they continue to encounter limitations.To address these issues, we propose GeoJSON Agents-a multi-agent LLM architecture.This framework transforms natural language tasks into structured GeoJSON operation commands and processes spatial data using two widely adopted LLM enhancement techniques:Function Calling and Code Generation.The architecture consists of three components-task parsing,agent collaboration,and result integration-aimed at enhancing both the performance and scalability of GIS automation.The Planner agent interprets natural language tasks into structured GeoJSON commands.Then,specialized Worker agents collaborate according to assigned roles to perform spatial data processing and analysis,either by invoking predefined function APIs or by dynamically generating and executing Python-based spatial analysis code.Finally,the system integrates the outputs from multiple execution rounds into reusable,standards-compliant GeoJSON files.To systematically evaluate the performance of the two approaches,we constructed a benchmark dataset of 70 tasks with varying complexity and conducted experiments using OpenAI's GPT-4o as the core model.Results indicate that the Function Calling-based GeoJSON Agent achieved an accuracy of 85.71%,while the Code Generation-based agent reached 97.14%,both significantly outperforming the best-performing general-purpose model (48.57%).Further analysis reveals that the Code Generation provides greater flexibility,whereas the Function Calling approach offers more stable execution.This study is the first to introduce an LLM multi-agent framework for GeoJSON data and to compare the strengths and limitations of two mainstream LLM enhancement methods,offering new perspectives for improving GeoAI system performance.
This study presents an Exploratory Retrieval-Augmented Planning (ExRAP) framework, designed to tackle continual instruction following tasks of embodied agents in dynamic, non-stationary environments. The framework enhances Large Language Models' (LLMs) embodied reasoning capabilities by efficiently exploring the physical environment and establishing the environmental context memory, thereby effectively grounding the task planning process in time-varying environment contexts. In ExRAP, given multiple continual instruction following tasks, each instruction is decomposed into queries on the environmental context memory and task executions conditioned on the query results. To efficiently handle these multiple tasks that are performed continuously and simultaneously, we implement an exploration-integrated task planning scheme by incorporating the {information-based exploration} into the LLM-based planning process. Combined with memory-augmented query evaluation, this integrated scheme not only allows for a better balance between the validity of the environmental context memory and the load of environment exploration, but also improves overall task performance. Furthermore, we devise a {temporal consistency refinement} scheme for query evaluation to address the inherent decay of knowledge in the memory. Through experiments with VirtualHome, ALFRED, and CARLA, our approach demonstrates robustness against a variety of embodied instruction following scenarios involving different instruction scales and types, and non-stationarity degrees, and it consistently outperforms other state-of-the-art LLM-based task planning approaches in terms of both goal success rate and execution efficiency.
Large Language Models (LLMs) often produce monolithic text that is hard to edit in parts, which can slow down collaborative workflows. We present componentization, an approach that decomposes model outputs into modular, independently editable units while preserving context. We describe Modular and Adaptable Output Decomposition (MAOD), which segments responses into coherent components and maintains links among them, and we outline the Component-Based Response Architecture (CBRA) as one way to implement this idea. Our reference prototype, MAODchat, uses a microservices design with state-machine-based decomposition agents, vendor-agnostic model adapters, and real-time component manipulation with recomposition. In an exploratory study with four participants from academic, engineering, and product roles, we observed that component-level editing aligned with several common workflows and enabled iterative refinement and selective reuse. Participants also mentioned possible team workflows. Our contributions are: (1) a definition of componentization for transforming monolithic outputs into manipulable units, (2) CBRA and MAODchat as a prototype architecture, (3) preliminary observations from a small user study, (4) MAOD as an algorithmic sketch for semantic segmentation, and (5) example Agent-to-Agent protocols for automated decomposition. We view componentization as a promising direction for turning passive text consumption into more active, component-level collaboration.
Structured prompting with XML tags has emerged as an effective way to steer large language models (LLMs) toward parseable, schema-adherent outputs in real-world systems. We develop a logic-first treatment of XML prompting that unifies (i) grammar-constrained decoding, (ii) fixed-point semantics over lattices of hierarchical prompts, and (iii) convergent human-AI interaction loops. We formalize a complete lattice of XML trees under a refinement order and prove that monotone prompt-to-prompt operators admit least fixed points (Knaster-Tarski) that characterize steady-state protocols; under a task-aware contraction metric on trees, we further prove Banach-style convergence of iterative guidance. We instantiate these results with context-free grammars (CFGs) for XML schemas and show how constrained decoding guarantees well-formedness while preserving task performance. A set of multi-layer human-AI interaction recipes demonstrates practical deployment patterns, including multi-pass "plan $\to$ verify $\to$ revise" routines and agentic tool use. We provide mathematically complete proofs and tie our framework to recent advances in grammar-aligned decoding, chain-of-verification, and programmatic prompting.