Large language models (LLMs) are growingly extended to process multimodal data such as text and video simultaneously. Their remarkable performance in understanding what is shown in images is surpassing specialized neural networks (NNs) such as Yolo that is supporting only a well-formed but very limited vocabulary, ie., objects that they are able to detect. When being non-restricted, LLMs and in particular state-of-the-art vision language models (VLMs) show impressive performance to describe even complex traffic situations. This is making them potentially suitable components for automotive perception systems to support the understanding of complex traffic situations or edge case situation. However, LLMs and VLMs are prone to hallucination, which mean to either potentially not seeing traffic agents such as vulnerable road users who are present in a situation, or to seeing traffic agents who are not there in reality. While the latter is unwanted making an ADAS or autonomous driving systems (ADS) to unnecessarily slow down, the former could lead to disastrous decisions from an ADS. In our work, we are systematically assessing the performance of 3 state-of-the-art VLMs on a diverse subset of traffic situations sampled from the Waymo Open Dataset to support safety guardrails for capturing such hallucinations in VLM-supported perception systems. We observe that both, proprietary and open VLMs exhibit remarkable image understanding capabilities even paying thorough attention to fine details sometimes difficult to spot for us humans. However, they are also still prone to making up elements in their descriptions to date requiring hallucination detection strategies such as BetterCheck that we propose in our work.
Large Language Model (LLM)-based autonomous agents are expected to play a vital role in the evolution of 6G networks, by empowering real-time decision-making related to management and service provisioning to end-users. This shift facilitates the transition from a specialized intelligence approach, where artificial intelligence (AI) algorithms handle isolated tasks, to artificial general intelligence (AGI)-driven networks, where agents possess broader reasoning capabilities and can manage diverse network functions. In this paper, we introduce a novel agentic paradigm that combines LLMs with real-time optimization algorithms towards Trustworthy AI, defined as symbiotic agents. Optimizers at the LLM's input-level provide bounded uncertainty steering for numerically precise tasks, whereas output-level optimizers supervised by the LLM enable adaptive real-time control. We design and implement two novel agent types including: (i) Radio Access Network optimizers, and (ii) multi-agent negotiators for Service-Level Agreements (SLAs). We further propose an end-to-end architecture for AGI networks and evaluate it on a 5G testbed capturing channel fluctuations from moving vehicles. Results show that symbiotic agents reduce decision errors fivefold compared to standalone LLM-based agents, while smaller language models (SLM) achieve similar accuracy with a 99.9% reduction in GPU resource overhead and in near-real-time loops of 82 ms. A multi-agent demonstration for collaborative RAN on the real-world testbed highlights significant flexibility in service-level agreement and resource allocation, reducing RAN over-utilization by approximately 44%. Drawing on our findings and open-source implementations, we introduce the symbiotic paradigm as the foundation for next-generation, AGI-driven networks-systems designed to remain adaptable, efficient, and trustworthy even as LLMs advance.
Understanding socio-ecological systems requires insights from diverse stakeholder perspectives, which are often hard to access. To enable alternative, simulation-based exploration of different stakeholder perspectives, we develop the HoPeS (Human-Oriented Perspective Shifting) modelling framework. HoPeS employs agents powered by large language models (LLMs) to represent various stakeholders; users can step into the agent roles to experience perspectival differences. A simulation protocol serves as a "scaffold" to streamline multiple perspective-taking simulations, supporting users in reflecting on, transitioning between, and integrating across perspectives. A prototype system is developed to demonstrate HoPeS in the context of institutional dynamics and land use change, enabling both narrative-driven and numerical experiments. In an illustrative experiment, a user successively adopts the perspectives of a system observer and a researcher - a role that analyses data from the embedded land use model to inform evidence-based decision-making for other LLM agents representing various institutions. Despite the user's effort to recommend technically sound policies, discrepancies persist between the policy recommendation and implementation due to stakeholders' competing advocacies, mirroring real-world misalignment between researcher and policymaker perspectives. The user's reflection highlights the subjective feelings of frustration and disappointment as a researcher, especially due to the challenge of maintaining political neutrality while attempting to gain political influence. Despite this, the user exhibits high motivation to experiment with alternative narrative framing strategies, suggesting the system's potential in exploring different perspectives. Further system and protocol refinement are likely to enable new forms of interdisciplinary collaboration in socio-ecological simulations.
Multi-step agentic retrieval systems based on large language models (LLMs) have demonstrated remarkable performance in complex information search tasks. However, these systems still face significant challenges in practical applications, particularly in generating factually inconsistent intermediate queries and inefficient search trajectories, which can lead to reasoning deviations or redundant computations. To address these issues, we propose DynaSearcher, an innovative search agent enhanced by dynamic knowledge graphs and multi-reward reinforcement learning (RL). Specifically, our system leverages knowledge graphs as external structured knowledge to guide the search process by explicitly modeling entity relationships, thereby ensuring factual consistency in intermediate queries and mitigating biases from irrelevant information. Furthermore, we employ a multi-reward RL framework for fine-grained control over training objectives such as retrieval accuracy, efficiency, and response quality. This framework promotes the generation of high-quality intermediate queries and comprehensive final answers, while discouraging unnecessary exploration and minimizing information omissions or redundancy. Experimental results demonstrate that our approach achieves state-of-the-art answer accuracy on six multi-hop question answering datasets, matching frontier LLMs while using only small-scale models and limited computational resources. Furthermore, our approach demonstrates strong generalization and robustness across diverse retrieval environments and larger-scale models, highlighting its broad applicability.
This paper presents Compliance Brain Assistant (CBA), a conversational, agentic AI assistant designed to boost the efficiency of daily compliance tasks for personnel in enterprise environments. To strike a good balance between response quality and latency, we design a user query router that can intelligently choose between (i) FastTrack mode: to handle simple requests that only need additional relevant context retrieved from knowledge corpora; and (ii) FullAgentic mode: to handle complicated requests that need composite actions and tool invocations to proactively discover context across various compliance artifacts, and/or involving other APIs/models for accommodating requests. A typical example would be to start with a user query, use its description to find a specific entity and then use the entity's information to query other APIs for curating and enriching the final AI response. Our experimental evaluations compared CBA against an out-of-the-box LLM on various real-world privacy/compliance-related queries targeting various personas. We found that CBA substantially improved upon the vanilla LLM's performance on metrics such as average keyword match rate (83.7% vs. 41.7%) and LLM-judge pass rate (82.0% vs. 20.0%). We also compared metrics for the full routing-based design against the `fast-track only` and `full-agentic` modes and found that it had a better average match-rate and pass-rate while keeping the run-time approximately the same. This finding validated our hypothesis that the routing mechanism leads to a good trade-off between the two worlds.
Analyzing large, complex output datasets from Discrete Event Simulations (DES) of warehouse operations to identify bottlenecks and inefficiencies is a critical yet challenging task, often demanding significant manual effort or specialized analytical tools. Our framework integrates Knowledge Graphs (KGs) and Large Language Model (LLM)-based agents to analyze complex Discrete Event Simulation (DES) output data from warehouse operations. It transforms raw DES data into a semantically rich KG, capturing relationships between simulation events and entities. An LLM-based agent uses iterative reasoning, generating interdependent sub-questions. For each sub-question, it creates Cypher queries for KG interaction, extracts information, and self-reflects to correct errors. This adaptive, iterative, and self-correcting process identifies operational issues mimicking human analysis. Our DES approach for warehouse bottleneck identification, tested with equipment breakdowns and process irregularities, outperforms baseline methods. For operational questions, it achieves near-perfect pass rates in pinpointing inefficiencies. For complex investigative questions, we demonstrate its superior diagnostic ability to uncover subtle, interconnected issues. This work bridges simulation modeling and AI (KG+LLM), offering a more intuitive method for actionable insights, reducing time-to-insight, and enabling automated warehouse inefficiency evaluation and diagnosis.
Central to agentic capability and trustworthiness of language model agents (LMAs) is the extent they maintain stable, reliable, identity over time. However, LMAs inherit pathologies from large language models (LLMs) (statelessness, stochasticity, sensitivity to prompts and linguistically-intermediation) which can undermine their identifiability, continuity, persistence and consistency. This attrition of identity can erode their reliability, trustworthiness and utility by interfering with their agentic capabilities such as reasoning, planning and action. To address these challenges, we introduce \textit{agent identity evals} (AIE), a rigorous, statistically-driven, empirical framework for measuring the degree to which an LMA system exhibit and maintain their agentic identity over time, including their capabilities, properties and ability to recover from state perturbations. AIE comprises a set of novel metrics which can integrate with other measures of performance, capability and agentic robustness to assist in the design of optimal LMA infrastructure and scaffolding such as memory and tools. We set out formal definitions and methods that can be applied at each stage of the LMA life-cycle, and worked examples of how to apply them.
This work tackles the physical layer security (PLS) problem of maximizing the secrecy rate in heterogeneous UAV networks (HetUAVNs) under propulsion energy constraints. Unlike prior studies that assume uniform UAV capabilities or overlook energy-security trade-offs, we consider a realistic scenario where UAVs with diverse payloads and computation resources collaborate to serve ground terminals in the presence of eavesdroppers. To manage the complex coupling between UAV motion and communication, we propose a hierarchical optimization framework. The inner layer uses a semidefinite relaxation (SDR)-based S2DC algorithm combining penalty functions and difference-of-convex (d.c.) programming to solve the secrecy precoding problem with fixed UAV positions. The outer layer introduces a Large Language Model (LLM)-guided heuristic multi-agent reinforcement learning approach (LLM-HeMARL) for trajectory optimization. LLM-HeMARL efficiently incorporates expert heuristics policy generated by the LLM, enabling UAVs to learn energy-aware, security-driven trajectories without the inference overhead of real-time LLM calls. The simulation results show that our method outperforms existing baselines in secrecy rate and energy efficiency, with consistent robustness across varying UAV swarm sizes and random seeds.
Role-Playing Language Agents (RPLAs) have emerged as a significant application direction for Large Language Models (LLMs). Existing approaches typically rely on prompt engineering or supervised fine-tuning to enable models to imitate character behaviors in specific scenarios, but often neglect the underlying \emph{cognitive} mechanisms driving these behaviors. Inspired by cognitive psychology, we introduce \textbf{CogDual}, a novel RPLA adopting a \textit{cognize-then-respond } reasoning paradigm. By jointly modeling external situational awareness and internal self-awareness, CogDual generates responses with improved character consistency and contextual alignment. To further optimize the performance, we employ reinforcement learning with two general-purpose reward schemes designed for open-domain text generation. Extensive experiments on the CoSER benchmark, as well as Cross-MR and LifeChoice, demonstrate that CogDual consistently outperforms existing baselines and generalizes effectively across diverse role-playing tasks.
Global health emergencies, such as the COVID-19 pandemic, have exposed critical weaknesses in traditional medical supply chains, including inefficiencies in resource allocation, lack of transparency, and poor adaptability to dynamic disruptions. This paper presents a novel hybrid framework that integrates blockchain technology with a decentralized, large language model (LLM) powered multi-agent negotiation system to enhance the resilience and accountability of medical supply chains during crises. In this system, autonomous agents-representing manufacturers, distributors, and healthcare institutions-engage in structured, context-aware negotiation and decision-making processes facilitated by LLMs, enabling rapid and ethical allocation of scarce medical resources. The off-chain agent layer supports adaptive reasoning and local decision-making, while the on-chain blockchain layer ensures immutable, transparent, and auditable enforcement of decisions via smart contracts. The framework also incorporates a formal cross-layer communication protocol to bridge decentralized negotiation with institutional enforcement. A simulation environment emulating pandemic scenarios evaluates the system's performance, demonstrating improvements in negotiation efficiency, fairness of allocation, supply chain responsiveness, and auditability. This research contributes an innovative approach that synergizes blockchain trust guarantees with the adaptive intelligence of LLM-driven agents, providing a robust and scalable solution for critical supply chain coordination under uncertainty.
Large language model (LLM) agents often struggle in environments where rules and required domain knowledge frequently change, such as regulatory compliance and user risk screening. Current approaches, like offline fine-tuning and standard prompting, are insufficient because they cannot effectively adapt to new knowledge during actual operation. To address this limitation, we propose the Adaptive Reflective Interactive Agent (ARIA), an LLM agent framework designed specifically to continuously learn updated domain knowledge at test time. ARIA assesses its own uncertainty through structured self-dialogue, proactively identifying knowledge gaps and requesting targeted explanations or corrections from human experts. It then systematically updates an internal, timestamped knowledge repository with provided human guidance, detecting and resolving conflicting or outdated knowledge through comparisons and clarification queries. We evaluate ARIA on the realistic customer due diligence name screening task on TikTok Pay, alongside publicly available dynamic knowledge tasks. Results demonstrate significant improvements in adaptability and accuracy compared to baselines using standard offline fine-tuning and existing self-improving agents. ARIA is deployed within TikTok Pay serving over 150 million monthly active users, confirming its practicality and effectiveness for operational use in rapidly evolving environments.
Multimodal learning plays a critical role in e-commerce recommendation platforms today, enabling accurate recommendations and product understanding. However, existing vision-language models, such as CLIP, face key challenges in e-commerce recommendation systems: 1) Weak object-level alignment, where global image embeddings fail to capture fine-grained product attributes, leading to suboptimal retrieval performance; 2) Ambiguous textual representations, where product descriptions often lack contextual clarity, affecting cross-modal matching; and 3) Domain mismatch, as generic vision-language models may not generalize well to e-commerce-specific data. To address these limitations, we propose a framework, VL-CLIP, that enhances CLIP embeddings by integrating Visual Grounding for fine-grained visual understanding and an LLM-based agent for generating enriched text embeddings. Visual Grounding refines image representations by localizing key products, while the LLM agent enhances textual features by disambiguating product descriptions. Our approach significantly improves retrieval accuracy, multimodal retrieval effectiveness, and recommendation quality across tens of millions of items on one of the largest e-commerce platforms in the U.S., increasing CTR by 18.6%, ATC by 15.5%, and GMV by 4.0%. Additional experimental results show that our framework outperforms vision-language models, including CLIP, FashionCLIP, and GCL, in both precision and semantic alignment, demonstrating the potential of combining object-aware visual grounding and LLM-enhanced text representation for robust multimodal recommendations.
Large language model (LLM) agents have shown increasing promise for collaborative task completion. However, existing multi-agent frameworks often rely on static workflows, fixed roles, and limited inter-agent communication, reducing their effectiveness in open-ended, high-complexity domains. This paper proposes a coordination framework that enables adaptiveness through three core mechanisms: dynamic task routing, bidirectional feedback, and parallel agent evaluation. The framework allows agents to reallocate tasks based on confidence and workload, exchange structured critiques to iteratively improve outputs, and crucially compete on high-ambiguity subtasks with evaluator-driven selection of the most suitable result. We instantiate these principles in a modular architecture and demonstrate substantial improvements in factual coverage, coherence, and efficiency over static and partially adaptive baselines. Our findings highlight the benefits of incorporating both adaptiveness and structured competition in multi-agent LLM systems.
Pairwise preferences over model responses are widely collected to evaluate and provide feedback to large language models (LLMs). Given two alternative model responses to the same input, a human or AI annotator selects the "better" response. This approach can provide feedback for domains where other hard-coded metrics are difficult to obtain (e.g., chat response quality), thereby helping model evaluation or training. However, for some domains high-quality pairwise comparisons can be tricky to obtain - from AI and humans. For example, for responses with many factual statements, annotators may disproportionately weigh writing quality rather than underlying facts. In this work, we explore augmenting standard AI annotator systems with additional tools to improve performance on three challenging response domains: long-form factual, math and code tasks. We propose a tool-using agentic system to provide higher quality feedback on these domains. Our system uses web-search and code execution to ground itself based on external validation, independent of the LLM's internal knowledge and biases. We provide extensive experimental results evaluating our method across the three targeted response domains as well as general annotation tasks, using RewardBench (incl. AlpacaEval and LLMBar), RewardMath, as well as three new datasets for domains with saturated pre-existing datasets. Our results indicate that external tools can indeed improve performance in many, but not all, cases. More generally, our experiments highlight the sensitivity of performance to simple parameters (e.g., prompt) and the need for improved (non-saturated) annotator benchmarks. We share our code at https://github.com/apple/ml-agent-evaluator.
Accessing knowledge via multilingual natural-language interfaces is one of the emerging challenges in the field of information retrieval and related ones. Structured knowledge stored in knowledge graphs can be queried via a specific query language (e.g., SPARQL). Therefore, one needs to transform natural-language input into a query to fulfill an information need. Prior approaches mostly focused on combining components (e.g., rule-based or neural-based) that solve downstream tasks and come up with an answer at the end. We introduce mKGQAgent, a human-inspired framework that breaks down the task of converting natural language questions into SPARQL queries into modular, interpretable subtasks. By leveraging a coordinated LLM agent workflow for planning, entity linking, and query refinement - guided by an experience pool for in-context learning - mKGQAgent efficiently handles multilingual KGQA. Evaluated on the DBpedia- and Corporate-based KGQA benchmarks within the Text2SPARQL challenge 2025, our approach took first place among the other participants. This work opens new avenues for developing human-like reasoning systems in multilingual semantic parsing.
Recent advancements in Large Language Models (LLMs) have catalyzed a paradigm shift from static prediction systems to agentic AI agents capable of reasoning, interacting with tools, and adapting to complex tasks. While LLM-based agentic systems have shown promise across many domains, their application to medical imaging remains in its infancy. In this work, we introduce AURA, the first visual linguistic explainability agent designed specifically for comprehensive analysis, explanation, and evaluation of medical images. By enabling dynamic interactions, contextual explanations, and hypothesis testing, AURA represents a significant advancement toward more transparent, adaptable, and clinically aligned AI systems. We highlight the promise of agentic AI in transforming medical image analysis from static predictions to interactive decision support. Leveraging Qwen-32B, an LLM-based architecture, AURA integrates a modular toolbox comprising: (i) a segmentation suite with phase grounding, pathology segmentation, and anatomy segmentation to localize clinically meaningful regions; (ii) a counterfactual image-generation module that supports reasoning through image-level explanations; and (iii) a set of evaluation tools including pixel-wise difference-map analysis, classification, and advanced state-of-the-art components to assess diagnostic relevance and visual interpretability.
Vision-language-action (VLA) reasoning tasks require agents to interpret multimodal instructions, perform long-horizon planning, and act adaptively in dynamic environments. Existing approaches typically train VLA models in an end-to-end fashion, directly mapping inputs to actions without explicit reasoning, which hinders their ability to plan over multiple steps or adapt to complex task variations. In this paper, we propose ThinkAct, a dual-system framework that bridges high-level reasoning with low-level action execution via reinforced visual latent planning. ThinkAct trains a multimodal LLM to generate embodied reasoning plans guided by reinforcing action-aligned visual rewards based on goal completion and trajectory consistency. These reasoning plans are compressed into a visual plan latent that conditions a downstream action model for robust action execution on target environments. Extensive experiments on embodied reasoning and robot manipulation benchmarks demonstrate that ThinkAct enables few-shot adaptation, long-horizon planning, and self-correction behaviors in complex embodied AI tasks.
We propose LingBench++, a linguistically-informed benchmark and reasoning framework designed to evaluate large language models (LLMs) on complex linguistic tasks inspired by the International Linguistics Olympiad (IOL). Unlike prior benchmarks that focus solely on final answer accuracy, LingBench++ provides structured reasoning traces, stepwise evaluation protocols, and rich typological metadata across over 90 low-resource and cross-cultural languages. We further develop a multi-agent architecture integrating grammatical knowledge retrieval, tool-augmented reasoning, and deliberate hypothesis testing. Through systematic comparisons of baseline and our proposed agentic models, we demonstrate that models equipped with external knowledge sources and iterative reasoning outperform single-pass approaches in both accuracy and interpretability. LingBench++ offers a comprehensive foundation for advancing linguistically grounded, culturally informed, and cognitively plausible reasoning in LLMs.
Large Language Models (LLMs) exhibit considerable promise in financial applications; however, prevailing models frequently demonstrate limitations when confronted with scenarios that necessitate sophisticated reasoning capabilities, stringent trustworthiness criteria, and efficient adaptation to domain-specific requirements. We introduce the Agentar-Fin-R1 series of financial large language models (8B and 32B parameters), specifically engineered based on the Qwen3 foundation model to enhance reasoning capabilities, reliability, and domain specialization for financial applications. Our optimization approach integrates a high-quality, systematic financial task label system with a comprehensive multi-layered trustworthiness assurance framework. This framework encompasses high-quality trustworthy knowledge engineering, multi-agent trustworthy data synthesis, and rigorous data validation governance. Through label-guided automated difficulty-aware optimization, tow-stage training pipeline, and dynamic attribution systems, we achieve substantial improvements in training efficiency. Our models undergo comprehensive evaluation on mainstream financial benchmarks including Fineva, FinEval, and FinanceIQ, as well as general reasoning datasets such as MATH-500 and GPQA-diamond. To thoroughly assess real-world deployment capabilities, we innovatively propose the Finova evaluation benchmark, which focuses on agent-level financial reasoning and compliance verification. Experimental results demonstrate that Agentar-Fin-R1 not only achieves state-of-the-art performance on financial tasks but also exhibits exceptional general reasoning capabilities, validating its effectiveness as a trustworthy solution for high-stakes financial applications. The Finova bench is available at https://github.com/antgroup/Finova.
The rapid advancement of large language models (LLMs) has enabled role-playing language agents to demonstrate significant potential in various applications. However, relying solely on prompts and contextual inputs often proves insufficient for achieving deep immersion in specific roles, particularly well-known fictional or public figures. On the other hand, fine-tuning-based approaches face limitations due to the challenges associated with data collection and the computational resources required for training, thereby restricting their broader applicability. To address these issues, we propose Test-Time-Matching (TTM), a training-free role-playing framework through test-time scaling and context engineering. TTM uses LLM agents to automatically decouple a character's features into personality, memory, and linguistic style. Our framework involves a structured, three-stage generation pipeline that utilizes these features for controlled role-playing. It achieves high-fidelity role-playing performance, also enables seamless combinations across diverse linguistic styles and even variations in personality and memory. We evaluate our framework through human assessment, and the results demonstrate that our method achieves the outstanding performance in generating expressive and stylistically consistent character dialogues.
Improving the reliability of large language models (LLMs) is critical for deploying them in real-world scenarios. In this paper, we propose \textbf{Deliberative Searcher}, the first framework to integrate certainty calibration with retrieval-based search for open-domain question answering. The agent performs multi-step reflection and verification over Wikipedia data and is trained with a reinforcement learning algorithm that optimizes for accuracy under a soft reliability constraint. Empirical results show that proposed method improves alignment between model confidence and correctness, leading to more trustworthy outputs. This paper will be continuously updated.
Agentic search, as a more autonomous and adaptive paradigm of retrieval augmentation, is driving the evolution of intelligent search systems. However, existing evaluation frameworks fail to align well with the goals of agentic search. First, the complex queries commonly used in current benchmarks often deviate from realistic user search scenarios. Second, prior approaches tend to introduce noise when extracting ground truth for end-to-end evaluations, leading to distorted assessments at a fine-grained level. Third, most current frameworks focus solely on the quality of final answers, neglecting the evaluation of the iterative process inherent to agentic search. To address these limitations, we propose RAVine -- a Reality-Aligned eValuation framework for agentic LLMs with search. RAVine targets multi-point queries and long-form answers that better reflect user intents, and introduces an attributable ground truth construction strategy to enhance the accuracy of fine-grained evaluation. Moreover, RAVine examines model's interaction with search tools throughout the iterative process, and accounts for factors of efficiency. We benchmark a series of models using RAVine and derive several insights, which we hope will contribute to advancing the development of agentic search systems. The code and datasets are available at https://github.com/SwordFaith/RAVine.
Conventional Retrieval-Augmented Generation (RAG) systems enhance Large Language Models (LLMs) but often fall short on complex queries, delivering limited, extractive answers and struggling with multiple targeted retrievals or navigating intricate entity relationships. This is a critical gap in knowledge-intensive domains. We introduce INRAExplorer, an agentic RAG system for exploring the scientific data of INRAE (France's National Research Institute for Agriculture, Food and Environment). INRAExplorer employs an LLM-based agent with a multi-tool architecture to dynamically engage a rich knowledge base, through a comprehensive knowledge graph derived from open access INRAE publications. This design empowers INRAExplorer to conduct iterative, targeted queries, retrieve exhaustive datasets (e.g., all publications by an author), perform multi-hop reasoning, and deliver structured, comprehensive answers. INRAExplorer serves as a concrete illustration of enhancing knowledge interaction in specialized fields.
Large Language Model (LLM) agents hold promise for a flexible and scalable alternative to traditional business process automation, but struggle to reliably follow complex company policies. In this study we introduce a deterministic, transparent, and modular framework for enforcing business policy adherence in agentic workflows. Our method operates in two phases: (1) an offline buildtime stage that compiles policy documents into verifiable guard code associated with tool use, and (2) a runtime integration where these guards ensure compliance before each agent action. We demonstrate our approach on the challenging $\tau$-bench Airlines domain, showing encouraging preliminary results in policy enforcement, and further outline key challenges for real-world deployments.
Atomic commits, each of which addresses a single development concern, are a best practice in software development. However, developers frequently produce tangled commits that mix unrelated changes due to practical constraints or unclear boundaries, negatively impacting code review and maintenance. Although prior commit untangling approaches: rule-based, feature-based, or graph-based, have made progress, they often rely on shallow signals and fail to distinguish between explicit dependencies (e.g., control/data flow) and implicit ones (e.g., semantic or conceptual relationships). In this paper, we propose ColaUntangle, a new collaborative consultation framework for commit untangling that models both explicit and implicit dependencies among code changes. ColaUntangle integrates Large Language Model (LLM)-driven agents in a multi-agent architecture: one agent specializes in explicit dependencies, another in implicit ones, and a reviewer agent synthesizes their perspectives through iterative consultation. To capture explicit and implicit contextual information, we construct multi-version Program Dependency Graphs (delta-PDG), enabling agents to reason over code relationships with both symbolic and semantic depth. We evaluate ColaUntangle on two widely-used datasets (1,612 C# and 14k Java tangled commits). Experimental results show that ColaUntangle outperforms the best-performing baseline, achieving an improvement of 44% on the C# dataset and 100% on the Java dataset. These findings highlight the potential of LLM-based collaborative frameworks for advancing automated commit untangling tasks.
Multi-Agent Systems (MAS) excel at accomplishing complex objectives through the collaborative efforts of individual agents. Among the methodologies employed in MAS, Multi-Agent Reinforcement Learning (MARL) stands out as one of the most efficacious algorithms. However, when confronted with the complex objective of Formation Control with Collision Avoidance (FCCA): designing an effective reward function that facilitates swift convergence of the policy network to an optimal solution. In this paper, we introduce a novel framework that aims to overcome this challenge. By giving large language models (LLMs) on the prioritization of tasks and the observable information available to each agent, our framework generates reward functions that can be dynamically adjusted online based on evaluation outcomes by employing more advanced evaluation metrics rather than the rewards themselves. This mechanism enables the MAS to simultaneously achieve formation control and obstacle avoidance in dynamic environments with enhanced efficiency, requiring fewer iterations to reach superior performance levels. Our empirical studies, conducted in both simulation and real-world settings, validate the practicality and effectiveness of our proposed approach.
The integration of voice-based AI agents in healthcare presents a transformative opportunity to bridge economic and accessibility gaps in digital health delivery. This paper explores the role of large language model (LLM)-powered voice assistants in enhancing preventive care and continuous patient monitoring, particularly in underserved populations. Drawing insights from the development and pilot study of Agent PULSE (Patient Understanding and Liaison Support Engine) -- a collaborative initiative between IBM Research, Cleveland Clinic Foundation, and Morehouse School of Medicine -- we present an economic model demonstrating how AI agents can provide cost-effective healthcare services where human intervention is economically unfeasible. Our pilot study with 33 inflammatory bowel disease patients revealed that 70\% expressed acceptance of AI-driven monitoring, with 37\% preferring it over traditional modalities. Technical challenges, including real-time conversational AI processing, integration with healthcare systems, and privacy compliance, are analyzed alongside policy considerations surrounding regulation, bias mitigation, and patient autonomy. Our findings suggest that AI-driven voice agents not only enhance healthcare scalability and efficiency but also improve patient engagement and accessibility. For healthcare executives, our cost-utility analysis demonstrates huge potential savings for routine monitoring tasks, while technologists can leverage our framework to prioritize improvements yielding the highest patient impact. By addressing current limitations and aligning AI development with ethical and regulatory frameworks, voice-based AI agents can serve as a critical entry point for equitable, sustainable digital healthcare solutions.
The automatic generation of Verilog code using Large Language Models (LLMs) has garnered significant interest in hardware design automation. However, existing benchmarks for evaluating LLMs in Verilog generation fall short in replicating real-world design workflows due to their designs' simplicity, inadequate design specifications, and less rigorous verification environments. To address these limitations, we present RealBench, the first benchmark aiming at real-world IP-level Verilog generation tasks. RealBench features complex, structured, real-world open-source IP designs, multi-modal and formatted design specifications, and rigorous verification environments, including 100% line coverage testbenches and a formal checker. It supports both module-level and system-level tasks, enabling comprehensive assessments of LLM capabilities. Evaluations on various LLMs and agents reveal that even one of the best-performing LLMs, o1-preview, achieves only a 13.3% pass@1 on module-level tasks and 0% on system-level tasks, highlighting the need for stronger Verilog generation models in the future. The benchmark is open-sourced at https://github.com/IPRC-DIP/RealBench.
Recent evidence suggests Large Language Models (LLMs) display Theory of Mind (ToM) abilities. Most ToM experiments place participants in a spectatorial role, wherein they predict and interpret other agents' behavior. However, human ToM also contributes to dynamically planning action and strategically intervening on others' mental states. We present MindGames: a novel `planning theory of mind' (PToM) task which requires agents to infer an interlocutor's beliefs and desires to persuade them to alter their behavior. Unlike previous evaluations, we explicitly evaluate use cases of ToM. We find that humans significantly outperform o1-preview (an LLM) at our PToM task (11% higher; $p=0.006$). We hypothesize this is because humans have an implicit causal model of other agents (e.g., they know, as our task requires, to ask about people's preferences). In contrast, o1-preview outperforms humans in a baseline condition which requires a similar amount of planning but minimal mental state inferences (e.g., o1-preview is better than humans at planning when already given someone's preferences). These results suggest a significant gap between human-like social reasoning and LLM abilities.
We report the discovery of a structural convergence across four influential theories of mind: Kahneman's dual-system theory, Friston's predictive processing, Minsky's society of mind, and Clark's extended mind-emerging unintentionally within a practical AI agent architecture called Agentic Flow. Designed to address limitations in large language models (LLMs), Agentic Flow comprises five interdependent modules such as Retrieval, Cognition, Control, Memory, and Action arranged in a recurrent cognitive loop. Although originally inspired only by Minsky and Clark, the system's structure retrospectively aligns with computational motifs found in all four theories, including predictive modeling, associative recall, and error-sensitive control. To assess this convergence, we conducted comparative experiments with baseline LLM agents on multi-step reasoning tasks. The structured agent achieved 95.8% task success and exhibited strong constraint adherence, while the baseline system succeeded 62.3% of the time. These results were not aimed at proving superiority, but at illustrating how theoretical structures may emerge through practical design choices rather than top-down theory. We introduce PEACE as a descriptive meta-architecture that captures design-level regularities observed in Agentic Flow. Not intended as a new theory, PEACE provides a shared vocabulary for understanding architectures shaped by real-world implementation demands. This paper should be read as a position paper - an exploratory reflection on how implementation can surface latent structural echoes of cognitive theory, without asserting theoretical unification.