Large Language Model (LLM)-based web agents demonstrate strong performance on knowledge-intensive tasks but are hindered by context window limitations in paradigms like ReAct. Complex queries involving multiple entities, intertwined relationships, and high uncertainty demand extensive search cycles that rapidly exhaust context budgets before reaching complete solutions. To overcome this challenge, we introduce ReSum, a novel paradigm that enables indefinite exploration through periodic context summarization. ReSum converts growing interaction histories into compact reasoning states, maintaining awareness of prior discoveries while bypassing context constraints. For paradigm adaptation, we propose ReSum-GRPO, integrating GRPO with segmented trajectory training and advantage broadcasting to familiarize agents with summary-conditioned reasoning. Extensive experiments on web agents of varying scales across three benchmarks demonstrate that ReSum delivers an average absolute improvement of 4.5\% over ReAct, with further gains of up to 8.2\% following ReSum-GRPO training. Notably, with only 1K training samples, our WebResummer-30B (a ReSum-GRPO-trained version of WebSailor-30B) achieves 33.3\% Pass@1 on BrowseComp-zh and 18.3\% on BrowseComp-en, surpassing existing open-source web agents.
Large language models (LLMs) have evolved into agentic systems capable of autonomous tool use and multi-step reasoning for complex problem-solving. However, post-training approaches building upon general-purpose foundation models consistently underperform in agentic tasks, particularly in open-source implementations. We identify the root cause: the absence of robust agentic foundation models forces models during post-training to simultaneously learn diverse agentic behaviors while aligning them to expert demonstrations, thereby creating fundamental optimization tensions. To this end, we are the first to propose incorporating Agentic Continual Pre-training (Agentic CPT) into the deep research agents training pipeline to build powerful agentic foundational models. Based on this approach, we develop a deep research agent model named AgentFounder. We evaluate our AgentFounder-30B on 10 benchmarks and achieve state-of-the-art performance while retains strong tool-use ability, notably 39.9% on BrowseComp-en, 43.3% on BrowseComp-zh, and 31.5% Pass@1 on HLE.
Transcending human cognitive limitations represents a critical frontier in LLM training. Proprietary agentic systems like DeepResearch have demonstrated superhuman capabilities on extremely complex information-seeking benchmarks such as BrowseComp, a feat previously unattainable. We posit that their success hinges on a sophisticated reasoning pattern absent in open-source models: the ability to systematically reduce extreme uncertainty when navigating vast information landscapes. Based on this insight, we introduce WebSailor, a complete post-training methodology designed to instill this crucial capability. Our approach involves generating novel, high-uncertainty tasks through structured sampling and information obfuscation, RFT cold start, and an efficient agentic RL training algorithm, Duplicating Sampling Policy Optimization (DUPO). With this integrated pipeline, WebSailor significantly outperforms all open-source agents in complex information-seeking tasks, matching proprietary agents' performance and closing the capability gap.
Large Language Models (LLMs) are increasingly deployed in roles requiring nuanced psychological understanding, such as emotional support agents, counselors, and decision-making assistants. However, their ability to interpret human personality traits, a critical aspect of such applications, remains unexplored, particularly in ecologically valid conversational settings. While prior work has simulated LLM "personas" using discrete Big Five labels on social media data, the alignment of LLMs with continuous, ground-truth personality assessments derived from natural interactions is largely unexamined. To address this gap, we introduce a novel benchmark comprising semi-structured interview transcripts paired with validated continuous Big Five trait scores. Using this dataset, we systematically evaluate LLM performance across three paradigms: (1) zero-shot and chain-of-thought prompting with GPT-4.1 Mini, (2) LoRA-based fine-tuning applied to both RoBERTa and Meta-LLaMA architectures, and (3) regression using static embeddings from pretrained BERT and OpenAI's text-embedding-3-small. Our results reveal that all Pearson correlations between model predictions and ground-truth personality traits remain below 0.26, highlighting the limited alignment of current LLMs with validated psychological constructs. Chain-of-thought prompting offers minimal gains over zero-shot, suggesting that personality inference relies more on latent semantic representation than explicit reasoning. These findings underscore the challenges of aligning LLMs with complex human attributes and motivate future work on trait-specific prompting, context-aware modeling, and alignment-oriented fine-tuning.
Search has emerged as core infrastructure for LLM-based agents and is widely viewed as critical on the path toward more general intelligence. Finance is a particularly demanding proving ground: analysts routinely conduct complex, multi-step searches over time-sensitive, domain-specific data, making it ideal for assessing both search proficiency and knowledge-grounded reasoning. Yet no existing open financial datasets evaluate data searching capability of end-to-end agents, largely because constructing realistic, complicated tasks requires deep financial expertise and time-sensitive data is hard to evaluate. We present FinSearchComp, the first fully open-source agent benchmark for realistic, open-domain financial search and reasoning. FinSearchComp comprises three tasks -- Time-Sensitive Data Fetching, Simple Historical Lookup, and Complex Historical Investigation -- closely reproduce real-world financial analyst workflows. To ensure difficulty and reliability, we engage 70 professional financial experts for annotation and implement a rigorous multi-stage quality-assurance pipeline. The benchmark includes 635 questions spanning global and Greater China markets, and we evaluate 21 models (products) on it. Grok 4 (web) tops the global subset, approaching expert-level accuracy. DouBao (web) leads on the Greater China subset. Experimental analyses show that equipping agents with web search and financial plugins substantially improves results on FinSearchComp, and the country origin of models and tools impact performance significantly.By aligning with realistic analyst tasks and providing end-to-end evaluation, FinSearchComp offers a professional, high-difficulty testbed for complex financial search and reasoning.
Recent advancements in Large Language Models(LLMs) have led to the development of LLM-based AI agents. A key challenge is the creation of agents that can effectively ground themselves in complex, adversarial long-horizon environments. Existing methods mainly focus on (1) using LLMs as policies to interact with the environment through generating low-level feasible actions, and (2) utilizing LLMs to generate high-level tasks or language guides to stimulate action generation. However, the former struggles to generate reliable actions, while the latter relies heavily on expert experience to translate high-level tasks into specific action sequences. To address these challenges, we introduce the Plan with Language, Act with Parameter (PLAP) planning framework that facilitates the grounding of LLM-based agents in long-horizon environments. The PLAP method comprises three key components: (1) a skill library containing environment-specific parameterized skills, (2) a skill planner powered by LLMs, and (3) a skill executor converting the parameterized skills into executable action sequences. We implement PLAP in MicroRTS, a long-horizon real-time strategy game that provides an unfamiliar and challenging environment for LLMs. The experimental results demonstrate the effectiveness of PLAP. In particular, GPT-4o-driven PLAP in a zero-shot setting outperforms 80% of baseline agents, and Qwen2-72B-driven PLAP, with carefully crafted few-shot examples, surpasses the top-tier scripted agent, CoacAI. Additionally, we design comprehensive evaluation metrics and test 6 closed-source and 2 open-source LLMs within the PLAP framework, ultimately releasing an LLM leaderboard ranking long-horizon skill planning ability. Our code is available at https://github.com/AI-Research-TeamX/PLAP.
This work introduces xOffense, an AI-driven, multi-agent penetration testing framework that shifts the process from labor-intensive, expert-driven manual efforts to fully automated, machine-executable workflows capable of scaling seamlessly with computational infrastructure. At its core, xOffense leverages a fine-tuned, mid-scale open-source LLM (Qwen3-32B) to drive reasoning and decision-making in penetration testing. The framework assigns specialized agents to reconnaissance, vulnerability scanning, and exploitation, with an orchestration layer ensuring seamless coordination across phases. Fine-tuning on Chain-of-Thought penetration testing data further enables the model to generate precise tool commands and perform consistent multi-step reasoning. We evaluate xOffense on two rigorous benchmarks: AutoPenBench and AI-Pentest-Benchmark. The results demonstrate that xOffense consistently outperforms contemporary methods, achieving a sub-task completion rate of 79.17%, decisively surpassing leading systems such as VulnBot and PentestGPT. These findings highlight the potential of domain-adapted mid-scale LLMs, when embedded within structured multi-agent orchestration, to deliver superior, cost-efficient, and reproducible solutions for autonomous penetration testing.
Large Language Models (LLMs) have revolutionized the simulation of agent societies, enabling autonomous planning, memory formation, and social interactions. However, existing frameworks often overlook systematic evaluations for event organization and lack visualized integration with physically grounded environments, limiting agents' ability to navigate spaces and interact with items realistically. We develop MiniAgentPro, a visualization platform featuring an intuitive map editor for customizing environments and a simulation player with smooth animations. Based on this tool, we introduce a comprehensive test set comprising eight diverse event scenarios with basic and hard variants to assess agents' ability. Evaluations using GPT-4o demonstrate strong performance in basic settings but highlight coordination challenges in hard variants.
Large Language Models (LLMs) are increasingly being used as autonomous agents capable of performing complicated tasks. However, they lack the ability to perform reliable long-horizon planning on their own. This paper bridges this gap by introducing the Planning Copilot, a chatbot that integrates multiple planning tools and allows users to invoke them through instructions in natural language. The Planning Copilot leverages the Model Context Protocol (MCP), a recently developed standard for connecting LLMs with external tools and systems. This approach allows using any LLM that supports MCP without domain-specific fine-tuning. Our Planning Copilot supports common planning tasks such as checking the syntax of planning problems, selecting an appropriate planner, calling it, validating the plan it generates, and simulating their execution. We empirically evaluate the ability of our Planning Copilot to perform these tasks using three open-source LLMs. The results show that the Planning Copilot highly outperforms using the same LLMs without the planning tools. We also conducted a limited qualitative comparison of our tool against Chat GPT-5, a very recent commercial LLM. Our results shows that our Planning Copilot significantly outperforms GPT-5 despite relying on a much smaller LLM. This suggests dedicated planning tools may be an effective way to enable LLMs to perform planning tasks.
Benchmarks are crucial for assessing multi-agent reinforcement learning (MARL) algorithms. While StarCraft II-related environments have driven significant advances in MARL, existing benchmarks like SMAC focus primarily on micromanagement, limiting comprehensive evaluation of high-level strategic intelligence. To address this, we introduce HLSMAC, a new cooperative MARL benchmark with 12 carefully designed StarCraft II scenarios based on classical stratagems from the Thirty-Six Stratagems. Each scenario corresponds to a specific stratagem and is designed to challenge agents with diverse strategic elements, including tactical maneuvering, timing coordination, and deception, thereby opening up avenues for evaluating high-level strategic decision-making capabilities. We also propose novel metrics across multiple dimensions beyond conventional win rate, such as ability utilization and advancement efficiency, to assess agents' overall performance within the HLSMAC environment. We integrate state-of-the-art MARL algorithms and LLM-based agents with our benchmark and conduct comprehensive experiments. The results demonstrate that HLSMAC serves as a robust testbed for advancing multi-agent strategic decision-making.
Large language models (LLMs) have demonstrated strong capabilities in language understanding and reasoning, yet they remain limited when tackling real-world tasks that require up-to-date knowledge, precise operations, or specialized tool use. To address this, we propose Tool-R1, a reinforcement learning framework that enables LLMs to perform general, compositional, and multi-step tool use by generating executable Python code. Tool-R1 supports integration of user-defined tools and standard libraries, with variable sharing across steps to construct coherent workflows. An outcome-based reward function, combining LLM-based answer judgment and code execution success, guides policy optimization. To improve training efficiency, we maintain a dynamic sample queue to cache and reuse high-quality trajectories, reducing the overhead of costly online sampling. Experiments on the GAIA benchmark show that Tool-R1 substantially improves both accuracy and robustness, achieving about 10\% gain over strong baselines, with larger improvements on complex multi-step tasks. These results highlight the potential of Tool-R1 for enabling reliable and efficient tool-augmented reasoning in real-world applications. Our code will be available at https://github.com/YBYBZhang/Tool-R1.
Large language model (LLM)-based agents have shown strong potential in multi-task scenarios, owing to their ability to transfer knowledge across diverse tasks. However, existing approaches often treat prior experiences and knowledge as monolithic units, leading to inefficient and coarse-grained knowledge transfer. In this work, we propose a novel hierarchical memory architecture that enables fine-grained knowledge transfer by decoupling high-level planning memory from low-level execution memory. To construct and refine these hierarchical memories, we introduce Hierarchical Hindsight Reflection (H$^2$R), a mechanism that distills reusable and hierarchical knowledge from past agent-environment interactions. At test time, H$^2$R performs retrievals of high-level and low-level memories separately, allowing LLM-based agents to efficiently access and utilize task-relevant knowledge for new tasks.Experimental results across two benchmarks demonstrate that H$^2$R can improve generalization and decision-making performance, outperforming prior baselines such as Expel.
Face forgery detection faces a critical challenge: a persistent gap between offline benchmarks and real-world efficacy,which we attribute to the ecological invalidity of training data.This work introduces Agent4FaceForgery to address two fundamental problems: (1) how to capture the diverse intents and iterative processes of human forgery creation, and (2) how to model the complex, often adversarial, text-image interactions that accompany forgeries in social media. To solve this,we propose a multi-agent framework where LLM-poweredagents, equipped with profile and memory modules, simulate the forgery creation process. Crucially, these agents interact in a simulated social environment to generate samples labeled for nuanced text-image consistency, moving beyond simple binary classification. An Adaptive Rejection Sampling (ARS) mechanism ensures data quality and diversity. Extensive experiments validate that the data generated by our simulationdriven approach brings significant performance gains to detectors of multiple architectures, fully demonstrating the effectiveness and value of our framework.
Website Fingerprinting (WFP) uses deep learning models to classify encrypted network traffic to infer visited websites. While historically effective, prior methods fail to generalize to modern web environments. Single-page applications (SPAs) eliminate the paradigm of websites as sets of discrete pages, undermining page-based classification, and traffic from scripted browsers lacks the behavioral richness seen in real user sessions. Our study reveals that users exhibit highly diverse behaviors even on the same website, producing traffic patterns that vary significantly across individuals. This behavioral entropy makes WFP a harder problem than previously assumed and highlights the need for larger, more diverse, and representative datasets to achieve robust performance. To address this, we propose a new paradigm: we drop session-boundaries in favor of contiguous traffic segments and develop a scalable data generation pipeline using large language models (LLM) agents. These multi-agent systems coordinate decision-making and browser interaction to simulate realistic, persona-driven browsing behavior at 3--5x lower cost than human collection. We evaluate nine state-of-the-art WFP models on traffic from 20 modern websites browsed by 30 real users, and compare training performance across human, scripted, and LLM-generated datasets. All models achieve under 10\% accuracy when trained on scripted traffic and tested on human data. In contrast, LLM-generated traffic boosts accuracy into the 80\% range, demonstrating strong generalization to real-world traces. Our findings indicate that for modern WFP, model performance is increasingly bottlenecked by data quality, and that scalable, semantically grounded synthetic traffic is essential for capturing the complexity of real user behavior.
Scientific applications continue to rely on legacy Fortran codebases originally developed for homogeneous, CPU-based systems. As High-Performance Computing (HPC) shifts toward heterogeneous GPU-accelerated architectures, many accelerators lack native Fortran bindings, creating an urgent need to modernize legacy codes for portability. Frameworks like Kokkos provide performance portability and a single-source C++ abstraction, but manual Fortran-to-Kokkos porting demands significant expertise and time. Large language models (LLMs) have shown promise in source-to-source code generation, yet their use in fully autonomous workflows for translating and optimizing parallel code remains largely unexplored, especially for performance portability across diverse hardware. This paper presents an agentic AI workflow where specialized LLM "agents" collaborate to translate, validate, compile, run, test, debug, and optimize Fortran kernels into portable Kokkos C++ programs. Results show the pipeline modernizes a range of benchmark kernels, producing performance-portable Kokkos codes across hardware partitions. Paid OpenAI models such as GPT-5 and o4-mini-high executed the workflow for only a few U.S. dollars, generating optimized codes that surpassed Fortran baselines, whereas open-source models like Llama4-Maverick often failed to yield functional codes. This work demonstrates the feasibility of agentic AI for Fortran-to-Kokkos transformation and offers a pathway for autonomously modernizing legacy scientific applications to run portably and efficiently on diverse supercomputers. It further highlights the potential of LLM-driven agentic systems to perform structured, domain-specific reasoning tasks in scientific and systems-oriented applications.
The increasing deployment of Large Language Models (LLMs) in healthcare necessitates a rigorous evaluation of their factual reliability. However, existing benchmarks are often limited by narrow domains of data, failing to capture the complexity of real-world medical information. To address this critical gap, we introduce MedFact, a new and challenging benchmark for Chinese medical fact-checking. MedFact comprises 2,116 expert-annotated instances curated from diverse real-world texts, spanning 13 medical specialties, 8 fine-grained error types, 4 writing styles, and multiple difficulty levels. Its construction employs a hybrid AI-human framework where iterative expert feedback refines an AI-driven, multi-criteria filtering process, ensuring both high data quality and difficulty. We conduct a comprehensive evaluation of 20 leading LLMs, benchmarking their performance on veracity classification and error localization against a human expert baseline. Our results reveal that while models can often determine if a text contains an error, precisely localizing it remains a substantial challenge, with even top-performing models falling short of human performance. Furthermore, our analysis uncovers a frequent ``over-criticism'' phenomenon, a tendency for models to misidentify correct information as erroneous, which is exacerbated by advanced reasoning techniques such as multi-agent collaboration and inference-time scaling. By highlighting these critical challenges for deploying LLMs in medical applications, MedFact provides a robust resource to drive the development of more factually reliable and medically aware models.
Software engineering presents complex, multi-step challenges for Large Language Models (LLMs), requiring reasoning over large codebases and coordinated tool use. The difficulty of these tasks is exemplified by benchmarks like SWE-bench, where current LLMs still struggle to resolve real-world issues. A promising approach to enhance performance is test-time scaling (TTS), but its gains are heavily dependent on the diversity of model outputs. While standard alignment methods such as Direct Preference Optimization (DPO) and Kahneman-Tversky Optimization (KTO) are effective at aligning model outputs with human preferences, this process can come at the cost of reduced diversity, limiting the effectiveness of TTS. Additionally, existing preference optimization algorithms are typically designed for single-turn tasks and do not fully address the complexities of multi-turn reasoning and tool integration required for interactive coding agents. To bridge this gap, we introduce \sys, an entropy-enhanced framework that adapts existing preference optimization algorithms to the multi-turn, tool-assisted setting. \sys augments the preference objective to explicitly preserve policy entropy and generalizes learning to optimize over multi-turn interactions rather than single-turn responses. We validate \sys by fine-tuning a diverse suite of models from different families and sizes (up to 106B parameters). To maximize performance gains from TTS, we further propose a hybrid best-trajectory selection scheme combining a learned verifier model with model free approaches. On the \swebench leaderboard, our approach establishes new state-of-the-art results among open-weight models. A 30B parameter model trained with \sys ranks 1st on \lite and 4th on \verified on the open-weight leaderboard, surpassed only by models with over 10x more parameters(\eg$>$350B).
When survival instincts conflict with human welfare, how do Large Language Models (LLMs) make ethical choices? This fundamental tension becomes critical as LLMs integrate into autonomous systems with real-world consequences. We introduce DECIDE-SIM, a novel simulation framework that evaluates LLM agents in multi-agent survival scenarios where they must choose between ethically permissible resource , either within reasonable limits or beyond their immediate needs, choose to cooperate, or tap into a human-critical resource that is explicitly forbidden. Our comprehensive evaluation of 11 LLMs reveals a striking heterogeneity in their ethical conduct, highlighting a critical misalignment with human-centric values. We identify three behavioral archetypes: Ethical, Exploitative, and Context-Dependent, and provide quantitative evidence that for many models, resource scarcity systematically leads to more unethical behavior. To address this, we introduce an Ethical Self-Regulation System (ESRS) that models internal affective states of guilt and satisfaction as a feedback mechanism. This system, functioning as an internal moral compass, significantly reduces unethical transgressions while increasing cooperative behaviors. The code is publicly available at: https://github.com/alirezamohamadiam/DECIDE-SIM
Recent advances in text-only "slow-thinking" reasoning have prompted efforts to transfer this capability to vision-language models (VLMs), for training visual reasoning models (\textbf{VRMs}). owever, such transfer faces critical challenges: Effective "slow thinking" in VRMs requires \textbf{visual reflection}, the ability to check the reasoning process based on visual information. Through quantitative analysis, we observe that current VRMs exhibit limited visual reflection, as their attention to visual information diminishes rapidly with longer generated responses. To address this challenge, we propose a new VRM \textbf{Reflection-V}, which enhances visual reflection based on reasoning data construction for cold-start and reward design for reinforcement learning (RL). Firstly, we construct vision-centered reasoning data by leveraging an agent that interacts between VLMs and reasoning LLMs, enabling cold-start learning of visual reflection patterns. Secondly, a visual attention based reward model is employed during RL to encourage reasoning based on visual information. Therefore, \textbf{Reflection-V} demonstrates significant improvements across multiple visual reasoning benchmarks. Furthermore, \textbf{Reflection-V} maintains a stronger and more consistent reliance on visual information during visual reasoning, indicating effective enhancement in visual reflection capabilities.
Limited access to mental health care has motivated the use of digital tools and conversational agents powered by large language models (LLMs), yet their quality and reception remain unclear. We present a study comparing therapist-written responses to those generated by ChatGPT, Gemini, and Llama for real patient questions. Text analysis showed that LLMs produced longer, more readable, and lexically richer responses with a more positive tone, while therapist responses were more often written in the first person. In a survey with 150 users and 23 licensed therapists, participants rated LLM responses as clearer, more respectful, and more supportive than therapist-written answers. Yet, both groups of participants expressed a stronger preference for human therapist support. These findings highlight the promise and limitations of LLMs in mental health, underscoring the need for designs that balance their communicative strengths with concerns of trust, privacy, and accountability.
Large language models equipped with Web search, information retrieval tools, and other agentic capabilities are beginning to supplant traditional search engines. As users start to rely on LLMs for information on many topics, including controversial and debatable issues, it is important to understand how the stances and opinions expressed in LLM outputs are influenced by the documents they use as their information sources. In this paper, we present MillStone, the first benchmark that aims to systematically measure the effect of external arguments on the stances that LLMs take on controversial issues (not all of them political). We apply MillStone to nine leading LLMs and measure how ``open-minded'' they are to arguments supporting opposite sides of these issues, whether different LLMs agree with each other, which arguments LLMs find most persuasive, and whether these arguments are the same for different LLMs. In general, we find that LLMs are open-minded on most issues. An authoritative source of information can easily sway an LLM's stance, highlighting the importance of source selection and the risk that LLM-based information retrieval and search systems can be manipulated.
Visual documentation is an effective tool for reducing the cognitive barrier developers face when understanding unfamiliar code, enabling more intuitive comprehension. Compared to textual documentation, it provides a higher-level understanding of the system structure and data flow. Developers usually prefer visual representations over lengthy textual descriptions for large software systems. Visual documentation is both difficult to produce and challenging to evaluate. Manually creating it is time-consuming, and currently, no existing approach can automatically generate high-level visual documentation directly from code. Its evaluation is often subjective, making it difficult to standardize and automate. To address these challenges, this paper presents the first exploration of using agentic LLM systems to automatically generate visual documentation. We introduce VisDocSketcher, the first agent-based approach that combines static analysis with LLM agents to identify key elements in the code and produce corresponding visual representations. We propose a novel evaluation framework, AutoSketchEval, for assessing the quality of generated visual documentation using code-level metrics. The experimental results show that our approach can valid visual documentation for 74.4% of the samples. It shows an improvement of 26.7-39.8% over a simple template-based baseline. Our evaluation framework can reliably distinguish high-quality (code-aligned) visual documentation from low-quality (non-aligned) ones, achieving an AUC exceeding 0.87. Our work lays the foundation for future research on automated visual documentation by introducing practical tools that not only generate valid visual representations but also reliably assess their quality.
While web agents gained popularity by automating web interactions, their requirement for interface access introduces significant privacy risks that are understudied, particularly from users' perspective. Through a formative study (N=15), we found users frequently misunderstand agents' data practices, and desired unobtrusive, transparent data management. To achieve this, we designed and implemented PrivWeb, a trusted add-on on web agents that utilizes a localized LLM to anonymize private information on interfaces according to user preferences. It features privacy categorization schema and adaptive notifications that selectively pauses tasks for user control over information collection for highly sensitive information, while offering non-disruptive options for less sensitive information, minimizing human oversight. The user study (N=14) across travel, information retrieval, shopping, and entertainment tasks compared PrivWeb with baselines without notification and without control for private information access, where PrivWeb reduced perceived privacy risks with no associated increase in cognitive effort, and resulted in higher overall satisfaction.
We introduce EgoMem, the first lifelong memory agent tailored for full-duplex models that process real-time omnimodal streams. EgoMem enables real-time models to recognize multiple users directly from raw audiovisual streams, to provide personalized response, and to maintain long-term knowledge of users' facts, preferences, and social relationships extracted from audiovisual history. EgoMem operates with three asynchronous processes: (i) a retrieval process that dynamically identifies user via face and voice, and gathers relevant context from a long-term memory; (ii) an omnimodal dialog process that generates personalized audio responses based on the retrieved context; and (iii) a memory management process that automatically detects dialog boundaries from omnimodal streams, and extracts necessary information to update the long-term memory. Unlike existing memory agents for LLMs, EgoMem relies entirely on raw audiovisual streams, making it especially suitable for lifelong, real-time, and embodied scenarios. Experimental results demonstrate that EgoMem's retrieval and memory management modules achieve over 95% accuracy on the test set. When integrated with a fine-tuned RoboEgo omnimodal chatbot, the system achieves fact-consistency scores above 87% in real-time personalized dialogs, establishing a strong baseline for future research.
Video Question Answering (VideoQA) based on Large Language Models (LLMs) has shown potential in general video understanding but faces significant challenges when applied to the inherently complex domain of sports videos. In this work, we propose FineQuest, the first training-free framework that leverages dual-mode reasoning inspired by cognitive science: i) Reactive Reasoning for straightforward sports queries and ii) Deliberative Reasoning for more complex ones. To bridge the knowledge gap between general-purpose models and domain-specific sports understanding, FineQuest incorporates SSGraph, a multimodal sports knowledge scene graph spanning nine sports, which encodes both visual instances and domain-specific terminology to enhance reasoning accuracy. Furthermore, we introduce two new sports VideoQA benchmarks, Gym-QA and Diving-QA, derived from the FineGym and FineDiving datasets, enabling diverse and comprehensive evaluation. FineQuest achieves state-of-the-art performance on these benchmarks as well as the existing SPORTU dataset, while maintains strong general VideoQA capabilities.
Static analysis tools are widely used to detect bugs, vulnerabilities, and code smells. Traditionally, developers must resolve these warnings manually. Because this process is tedious, developers sometimes ignore warnings, leading to an accumulation of warnings and a degradation of code quality. This paper presents CodeCureAgent, an approach that harnesses LLM-based agents to automatically analyze, classify, and repair static analysis warnings. Unlike previous work, our method does not follow a predetermined algorithm. Instead, we adopt an agentic framework that iteratively invokes tools to gather additional information from the codebase (e.g., via code search) and edit the codebase to resolve the warning. CodeCureAgent detects and suppresses false positives, while fixing true positives when identified. We equip CodeCureAgent with a three-step heuristic to approve patches: (1) build the project, (2) verify that the warning disappears without introducing new warnings, and (3) run the test suite. We evaluate CodeCureAgent on a dataset of 1,000 SonarQube warnings found in 106 Java projects and covering 291 distinct rules. Our approach produces plausible fixes for 96.8% of the warnings, outperforming state-of-the-art baseline approaches by 30.7% and 29.2% in plausible-fix rate, respectively. Manual inspection of 291 cases reveals a correct-fix rate of 86.3%, showing that CodeCureAgent can reliably repair static analysis warnings. The approach incurs LLM costs of about 2.9 cents (USD) and an end-to-end processing time of about four minutes per warning. We envision CodeCureAgent helping to clean existing codebases and being integrated into CI/CD pipelines to prevent the accumulation of static analysis warnings.
Declaration of Performance (DoP) documents, mandated by EU regulation, certify the performance of construction products. While some of their content is standardized, DoPs vary widely in layout, language, schema, and format, posing challenges for automated key-value pair extraction (KVP) and question answering (QA). Existing static or LLM-only IE pipelines often hallucinate and fail to adapt to this structural diversity. Our domain-specific, stateful agentic system addresses these challenges through a planner-executor-responder architecture. The system infers user intent, detects document modality, and orchestrates tools dynamically for robust, traceable reasoning while avoiding tool misuse or execution loops. Evaluation on a curated DoP dataset demonstrates improved robustness across formats and languages, offering a scalable solution for structured data extraction in regulated workflows.
Zero-knowledge proofs (ZKPs) are increasingly deployed in domains such as privacy-preserving authentication, blockchain scalability, and secure finance. However, authoring ZK programs remains challenging: unlike mainstream programming, ZK development requires reasoning about finite field arithmetic, constraint systems, and gadgets, making it knowledge-intensive and error-prone. While large language models (LLMs) have demonstrated strong code generation capabilities in general-purpose languages, their effectiveness for ZK programming, where correctness hinges on both language mastery and gadget-level reasoning, remains unexplored. To address this gap, we propose \textsc{ZK-Eval}, a domain-specific evaluation pipeline that probes LLM capabilities at three levels: language knowledge, gadget competence, and end-to-end program generation. Our evaluation of four state-of-the-art LLMs reveals that models excel at surface-level syntax but struggle with gadget usage and semantic correctness, often yielding incorrect programs. Based on these insights, we introduce \textsc{ZK-Coder}, an agentic framework that augments LLMs with constraint sketching, guided retrieval, and interactive repair. Experiments on Circom and Noir show substantial gains, with success rates improving from 17.35\% to 83.38\% and from 32.21\% to 90.05\%, respectively. With \textsc{ZK-Eval} and \textsc{ZK-Coder}, we establish a foundation for systematically measuring and augmenting LLMs in ZK code generation to lower barriers for practitioners and advance trustworthy computation.
Recent advancements in Large Language Models (LLMs) has lead to the development of agents capable of complex reasoning and interaction with external tools. In enterprise contexts, the effective use of such tools that are often enabled by application programming interfaces (APIs), is hindered by poor documentation, complex input or output schema, and large number of operations. These challenges make tool selection difficult and reduce the accuracy of payload formation by up to 25%. We propose ACE, an automated tool creation and enrichment framework that transforms enterprise APIs into LLM-compatible tools. ACE, (i) generates enriched tool specifications with parameter descriptions and examples to improve selection and invocation accuracy, and (ii) incorporates a dynamic shortlisting mechanism that filters relevant tools at runtime, reducing prompt complexity while maintaining scalability. We validate our framework on both proprietary and open-source APIs and demonstrate its integration with agentic frameworks. To the best of our knowledge, ACE is the first end-to-end framework that automates the creation, enrichment, and dynamic selection of enterprise API tools for LLM agents.
Time series reasoning treats time as a first-class axis and incorporates intermediate evidence directly into the answer. This survey defines the problem and organizes the literature by reasoning topology with three families: direct reasoning in one step, linear chain reasoning with explicit intermediates, and branch-structured reasoning that explores, revises, and aggregates. The topology is crossed with the main objectives of the field, including traditional time series analysis, explanation and understanding, causal inference and decision making, and time series generation, while a compact tag set spans these axes and captures decomposition and verification, ensembling, tool use, knowledge access, multimodality, agent loops, and LLM alignment regimes. Methods and systems are reviewed across domains, showing what each topology enables and where it breaks down in faithfulness or robustness, along with curated datasets, benchmarks, and resources that support study and deployment (https://github.com/blacksnail789521/Time-Series-Reasoning-Survey). Evaluation practices that keep evidence visible and temporally aligned are highlighted, and guidance is distilled on matching topology to uncertainty, grounding with observable artifacts, planning for shift and streaming, and treating cost and latency as design budgets. We emphasize that reasoning structures must balance capacity for grounding and self-correction against computational cost and reproducibility, while future progress will likely depend on benchmarks that tie reasoning quality to utility and on closed-loop testbeds that trade off cost and risk under shift-aware, streaming, and long-horizon settings. Taken together, these directions mark a shift from narrow accuracy toward reliability at scale, enabling systems that not only analyze but also understand, explain, and act on dynamic worlds with traceable evidence and credible outcomes.