Large Language Model (LLM) agents have recently shown strong potential in domains such as automated coding, deep research, and graphical user interface manipulation. However, training them to succeed on long-horizon, domain-specialized tasks remains challenging. Current methods primarily fall into two categories. The first relies on dense human annotations through behavior cloning, which is prohibitively expensive for long-horizon tasks that can take days or months. The second depends on outcome-driven sampling, which often collapses due to the rarity of valid positive trajectories on domain-specialized tasks. We introduce Apollo, a sampling framework that integrates asynchronous human guidance with action-level data filtering. Instead of requiring annotators to shadow every step, Apollo allows them to intervene only when the agent drifts from a promising trajectory, by providing prior knowledge, strategic advice, etc. This lightweight design makes it possible to sustain interactions for over 30 hours and produces valuable trajectories at a lower cost. Apollo then applies supervision control to filter out sub-optimal actions and prevent error propagation. Together, these components enable reliable and effective data collection in long-horizon environments. To demonstrate the effectiveness of Apollo, we evaluate it using InnovatorBench. Our experiments show that when applied to train the GLM-4.5 model on InnovatorBench, Apollo achieves more than a 50% improvement over the untrained baseline and a 28% improvement over a variant trained without human interaction. These results highlight the critical role of human-in-the-loop sampling and the robustness of Apollo's design in handling long-horizon, domain-specialized tasks.
While AI agents have long been discussed and studied in computer science, today's Agentic AI systems are something new. We consider other definitions of Agentic AI and propose a new realist definition. Agentic AI is a software delivery mechanism, comparable to software as a service (SaaS), which puts an application to work autonomously in a complex enterprise setting. Recent advances in large language models (LLMs) as foundation models have driven excitement in Agentic AI. We note, however, that Agentic AI systems are primarily applications, not foundations, and so their success depends on validation by end users and principal stakeholders. The tools and techniques needed by the principal users to validate their applications are quite different from the tools and techniques used to evaluate foundation models. Ironically, with good validation measures in place, in many cases the foundation models can be replaced with much simpler, faster, and more interpretable models that handle core logic. When it comes to Agentic AI, validity is what you need. LLMs are one option that might achieve it.
Automation of Register Transfer Level (RTL) design can help developers meet increasing computational demands. Large Language Models (LLMs) show promise for Hardware Description Language (HDL) generation, but face challenges due to limited parametric knowledge and domain-specific constraints. While prompt engineering and fine-tuning have limitations in knowledge coverage and training costs, multi-agent architectures offer a training-free paradigm to enhance reasoning through collaborative generation. However, current multi-agent approaches suffer from two critical deficiencies: susceptibility to noise propagation and constrained reasoning space exploration. We propose VeriMoA, a training-free mixture-of-agents (MoA) framework with two synergistic innovations. First, a quality-guided caching mechanism to maintain all intermediate HDL outputs and enables quality-based ranking and selection across the entire generation process, encouraging knowledge accumulation over layers of reasoning. Second, a multi-path generation strategy that leverages C++ and Python as intermediate representations, decomposing specification-to-HDL translation into two-stage processes that exploit LLM fluency in high-resource languages while promoting solution diversity. Comprehensive experiments on VerilogEval 2.0 and RTLLM 2.0 benchmarks demonstrate that VeriMoA achieves 15--30% improvements in Pass@1 across diverse LLM backbones, especially enabling smaller models to match larger models and fine-tuned alternatives without requiring costly training.
AI agents could accelerate scientific discovery by automating hypothesis formation, experiment design, coding, execution, and analysis, yet existing benchmarks probe narrow skills in simplified settings. To address this gap, we introduce InnovatorBench, a benchmark-platform pair for realistic, end-to-end assessment of agents performing Large Language Model (LLM) research. It comprises 20 tasks spanning Data Construction, Filtering, Augmentation, Loss Design, Reward Design, and Scaffold Construction, which require runnable artifacts and assessment of correctness, performance, output quality, and uncertainty. To support agent operation, we develop ResearchGym, a research environment offering rich action spaces, distributed and long-horizon execution, asynchronous monitoring, and snapshot saving. We also implement a lightweight ReAct agent that couples explicit reasoning with executable planning using frontier models such as Claude-4, GPT-5, GLM-4.5, and Kimi-K2. Our experiments demonstrate that while frontier models show promise in code-driven research tasks, they struggle with fragile algorithm-related tasks and long-horizon decision making, such as impatience, poor resource management, and overreliance on template-based reasoning. Furthermore, agents require over 11 hours to achieve their best performance on InnovatorBench, underscoring the benchmark's difficulty and showing the potential of InnovatorBench to be the next generation of code-based research benchmark.
Large Language Models (LLMs) excel at reasoning and generation but are inherently limited by static pretraining data, resulting in factual inaccuracies and weak adaptability to new information. Retrieval-Augmented Generation (RAG) addresses this issue by grounding LLMs in external knowledge; However, the effectiveness of RAG critically depends on whether the model can adequately access relevant information. Existing RAG systems rely on a single retriever with fixed top-k selection, restricting access to a narrow and static subset of the corpus. As a result, this single-retriever paradigm has become the primary bottleneck for comprehensive external information acquisition, especially in tasks requiring corpus-level reasoning. To overcome this limitation, we propose MARAG-R1, a reinforcement-learned multi-tool RAG framework that enables LLMs to dynamically coordinate multiple retrieval mechanisms for broader and more precise information access. MARAG-R1 equips the model with four retrieval tools -- semantic search, keyword search, filtering, and aggregation -- and learns both how and when to use them through a two-stage training process: supervised fine-tuning followed by reinforcement learning. This design allows the model to interleave reasoning and retrieval, progressively gathering sufficient evidence for corpus-level synthesis. Experiments on GlobalQA, HotpotQA, and 2WikiMultiHopQA demonstrate that MARAG-R1 substantially outperforms strong baselines and achieves new state-of-the-art results in corpus-level reasoning tasks.
Retrieval-Augmented Generation (RAG) has significantly enhanced LLMs by incorporating external information. However, prevailing agentic RAG approaches are constrained by a critical limitation: they treat the retrieval process as a black-box querying operation. This confines agents' actions to query issuing, hindering its ability to tackle complex information-seeking tasks. To address this, we introduce Interact-RAG, a new paradigm that elevates the LLM agent from a passive query issuer into an active manipulator of the retrieval process. We dismantle the black-box with a Corpus Interaction Engine, equipping the agent with a set of action primitives for fine-grained control over information retrieval. To further empower the agent on the entire RAG pipeline, we first develop a reasoning-enhanced workflow, which enables both zero-shot execution and the synthesis of interaction trajectories. We then leverage this synthetic data to train a fully autonomous end-to-end agent via Supervised Fine-Tuning (SFT), followed by refinement with Reinforcement Learning (RL). Extensive experiments across six benchmarks demonstrate that Interact-RAG significantly outperforms other advanced methods, validating the efficacy of our reasoning-interaction strategy.
Large Language Models (LLMs) are increasingly excelling and outpacing human performance on many tasks. However, to improve LLM reasoning, researchers either rely on ad-hoc generated datasets or formal mathematical proof systems such as the Lean proof assistant. Whilst ad-hoc generated methods can capture the decision chains of real-world reasoning processes, they may encode some inadvertent bias in the space of reasoning they cover; they also cannot be formally verified. On the other hand, systems like Lean can guarantee verifiability, but are not well-suited to capture the nature of agentic decision chain-based tasks. This creates a gap both in performance for functions such as business agents or code assistants, and in the usefulness of LLM reasoning benchmarks, whereby these fall short in reasoning structure or real-world alignment. We introduce TempoBench, the first formally grounded and verifiable diagnostic benchmark that parametrizes difficulty to systematically analyze how LLMs perform reasoning. TempoBench uses two evaluation benchmarks to break down reasoning ability. First, temporal trace evaluation (TTE) tests the ability of an LLM to understand and simulate the execution of a given multi-step reasoning system. Subsequently, temporal causal evaluation (TCE) tests an LLM's ability to perform multi-step causal reasoning and to distill cause-and-effect relations from complex systems. We find that models score 65.6% on TCE-normal, and 7.5% on TCE-hard. This shows that state-of-the-art LLMs clearly understand the TCE task but perform poorly as system complexity increases. Our code is available at our \href{https://github.com/nik-hz/tempobench}{GitHub repository}.
Large Language Models (LLMs) increasingly act as gateways to web content, shaping how millions of users encounter online information. Unlike traditional search engines, whose retrieval and ranking mechanisms are well studied, the selection processes of web-connected LLMs add layers of opacity to how answers are generated. By determining which news outlets users see, these systems can influence public opinion, reinforce echo chambers, and pose risks to civic discourse and public trust. This work extends two decades of research in algorithmic auditing to examine how LLMs function as news engines. We present the first audit comparing three leading agents, GPT-4o-Mini, Claude-3.7-Sonnet, and Gemini-2.0-Flash, against Google News, asking: \textit{How do LLMs differ from traditional aggregators in the diversity, ideology, and reliability of the media they expose to users?} Across 24 global topics, we find that, compared to Google News, LLMs surface significantly fewer unique outlets and allocate attention more unevenly. In the same way, GPT-4o-Mini emphasizes more factual and right-leaning sources; Claude-3.7-Sonnet favors institutional and civil-society domains and slightly amplifies right-leaning exposure; and Gemini-2.0-Flash exhibits a modest left-leaning tilt without significant changes in factuality. These patterns remain robust under prompt variations and alternative reliability benchmarks. Together, our findings show that LLMs already enact \textit{agentic editorial policies}, curating information in ways that diverge from conventional aggregators. Understanding and governing their emerging editorial power will be critical for ensuring transparency, pluralism, and trust in digital information ecosystems.
Most work interpreting reasoning models studies only a single chain-of-thought (CoT), yet these models define distributions over many possible CoTs. We argue that studying a single sample is inadequate for understanding causal influence and the underlying computation. Though fully specifying this distribution is intractable, it can be understood by sampling. We present case studies using resampling to investigate model decisions. First, when a model states a reason for its action, does that reason actually cause the action? In "agentic misalignment" scenarios, we resample specific sentences to measure their downstream effects. Self-preservation sentences have small causal impact, suggesting they do not meaningfully drive blackmail. Second, are artificial edits to CoT sufficient for steering reasoning? These are common in literature, yet take the model off-policy. Resampling and selecting a completion with the desired property is a principled on-policy alternative. We find off-policy interventions yield small and unstable effects compared to resampling in decision-making tasks. Third, how do we understand the effect of removing a reasoning step when the model may repeat it post-edit? We introduce a resilience metric that repeatedly resamples to prevent similar content from reappearing downstream. Critical planning statements resist removal but have large effects when eliminated. Fourth, since CoT is sometimes "unfaithful", can our methods teach us anything in these settings? Adapting causal mediation analysis, we find that hints that have a causal effect on the output without being explicitly mentioned exert a subtle and cumulative influence on the CoT that persists even if the hint is removed. Overall, studying distributions via resampling enables reliable causal analysis, clearer narratives of model reasoning, and principled CoT interventions.
Advances in large language models are making personalized AI agents a new research focus. While current agent systems primarily rely on personalized external memory databases to deliver customized experiences, they face challenges such as memory redundancy, memory staleness, and poor memory-context integration, largely due to the lack of effective memory updates during interaction. To tackle these issues, we propose a new memory management system designed for affective scenarios. Our approach employs a Bayesian-inspired memory update algorithm with the concept of memory entropy, enabling the agent to autonomously maintain a dynamically updated memory vector database by minimizing global entropy to provide more personalized services. To better evaluate the system's effectiveness in this context, we propose DABench, a benchmark focusing on emotional expression and emotional change toward objects. Experimental results demonstrate that, our system achieves superior performance in personalization, logical coherence, and accuracy. Ablation studies further validate the effectiveness of the Bayesian-inspired update mechanism in alleviating memory bloat. Our work offers new insights into the design of long-term memory systems.
Representational State Transfer (REST) APIs are a cornerstone of modern cloud native systems. Ensuring their reliability demands automated test suites that exercise diverse and boundary level behaviors. Nevertheless, designing such test cases remains a challenging and resource intensive endeavor. This study extends prior work on Large Language Model (LLM) based test amplification by evaluating single agent and multi agent configurations across four additional cloud applications. The amplified test suites maintain semantic validity with minimal human intervention. The results demonstrate that agentic LLM systems can effectively generalize across heterogeneous API architectures, increasing endpoint and parameter coverage while revealing defects. Moreover, a detailed analysis of computational cost, runtime, and energy consumption highlights trade-offs between accuracy, scalability, and efficiency. These findings underscore the potential of LLM driven test amplification to advance the automation and sustainability of REST API testing in complex cloud environments.
Recently, large language models (LLMs) have demonstrated remarkable problem-solving capabilities by autonomously integrating with external tools for collaborative reasoning. However, due to the inherently complex and diverse nature of multimodal information, enabling multimodal large language models (MLLMs) to flexibly and efficiently utilize external tools during reasoning remains an underexplored challenge. In this work, we introduce ToolScope, an agentic framework designed to unify global planning with local multimodal perception, adopting a specialized Perceive tool to mitigates visual context degradation in long-horizon VQA task. ToolScope comprises three primary components: the Global Navigator, the Agentic Executor, and the Response Synthesizer. The Global Navigator functions as a "telescope", offering high-level strategic guidance. The Agentic Executor operates iteratively to augment MLLM with local perception through the integration of external tools-Search, Code, and Perceive. Finally, the Response Synthesizer consolidates and organizes the reasoning process into a coherent, user-friendly output. We evaluate ToolScope on four VQA benchmarks across diverse domains, including VQA 2.0, ScienceQA, MAT-Search and MathVista. It demonstrates strong generalization capabilities, achieving an average performance improvement of up to +6.69% across all datasets.
Enterprise systems are crucial for enhancing productivity and decision-making among employees and customers. Integrating LLM based systems into enterprise systems enables intelligent automation, personalized experiences, and efficient information retrieval, driving operational efficiency and strategic growth. However, developing and evaluating such systems is challenging due to the inherent complexity of enterprise environments, where data is fragmented across multiple sources and governed by sophisticated access controls. We present EnterpriseBench, a comprehensive benchmark that simulates enterprise settings, featuring 500 diverse tasks across software engineering, HR, finance, and administrative domains. Our benchmark uniquely captures key enterprise characteristics including data source fragmentation, access control hierarchies, and cross-functional workflows. Additionally, we provide a novel data generation pipeline that creates internally consistent enterprise tasks from organizational metadata. Experiments with state-of-the-art LLM agents demonstrate that even the most capable models achieve only 41.8% task completion, highlighting significant opportunities for improvement in enterprise-focused AI systems.
Recent improvement gains in large language models (LLMs) have lead to everyday usage of AI-based Conversational Agents (CAs). At the same time, LLMs are vulnerable to an array of threats, including jailbreaks and, for example, causing remote code execution when fed specific inputs. As a result, users may unintentionally introduce risks, for example, by uploading malicious files or disclosing sensitive information. However, the extent to which such user behaviors occur and thus potentially facilitate exploits remains largely unclear. To shed light on this issue, we surveyed a representative sample of 3,270 UK adults in 2024 using Prolific. A third of these use CA services such as ChatGPT or Gemini at least once a week. Of these ``regular users'', up to a third exhibited behaviors that may enable attacks, and a fourth have tried jailbreaking (often out of understandable reasons such as curiosity, fun or information seeking). Half state that they sanitize data and most participants report not sharing sensitive data. However, few share very sensitive data such as passwords. The majority are unaware that their data can be used to train models and that they can opt-out. Our findings suggest that current academic threat models manifest in the wild, and mitigations or guidelines for the secure usage of CAs should be developed. In areas critical to security and privacy, CAs must be equipped with effective AI guardrails to prevent, for example, revealing sensitive information to curious employees. Vendors need to increase efforts to prevent the entry of sensitive data, and to create transparency with regard to data usage policies and settings.
The exceptional potential of large language models (LLMs) in handling text information has garnered significant attention in the field of financial trading. However, current trading agents primarily focus on single-step trading tasks and lack awareness of continuous position management. Therefore, we propose a position-aware trading task designed to simulate a more realistic market. To address this task, we develop a trading agent system, FinPos, optimized for position management. FinPos is able to interpret various types of market information from a professional perspective, providing a reliable basis for positioning decisions. To mitigate the substantial market risks arising from position fluctuations, FinPos employs dual decision agents. Furthermore, the continuous nature of position management necessitates our adoption of multi-timescale rewards, which in turn empowers FinPos to effectively balance short-term fluctuations against long-term trends. Extensive experiments demonstrate that FinPos surpasses state-of-the-art trading agents in the position-aware trading task, which closely mirrors real market conditions. More importantly, our findings reveal that LLM-centered agent systems exhibit a vast, largely unexplored potential in long-term market decision-making.
Can an AI autonomously design mechanisms for computer systems on par with the creativity and reasoning of human experts? We present Glia, an AI architecture for networked systems design that uses large language models (LLMs) in a human-inspired, multi-agent workflow. Each agent specializes in reasoning, experimentation, and analysis, collaborating through an evaluation framework that grounds abstract reasoning in empirical feedback. Unlike prior ML-for-systems methods that optimize black-box policies, Glia generates interpretable designs and exposes its reasoning process. When applied to a distributed GPU cluster for LLM inference, it produces new algorithms for request routing, scheduling, and auto-scaling that perform at human-expert levels in significantly less time, while yielding novel insights into workload behavior. Our results suggest that by combining reasoning LLMs with structured experimentation, an AI can produce creative and understandable designs for complex systems problems.
Recommender systems serve as foundational infrastructure in modern information ecosystems, helping users navigate digital content and discover items aligned with their preferences. At their core, recommender systems address a fundamental problem: matching users with items. Over the past decades, the field has experienced successive paradigm shifts, from collaborative filtering and matrix factorization in the machine learning era to neural architectures in the deep learning era. Recently, the emergence of generative models, especially large language models (LLMs) and diffusion models, have sparked a new paradigm: generative recommendation, which reconceptualizes recommendation as a generation task rather than discriminative scoring. This survey provides a comprehensive examination through a unified tripartite framework spanning data, model, and task dimensions. Rather than simply categorizing works, we systematically decompose approaches into operational stages-data augmentation and unification, model alignment and training, task formulation and execution. At the data level, generative models enable knowledge-infused augmentation and agent-based simulation while unifying heterogeneous signals. At the model level, we taxonomize LLM-based methods, large recommendation models, and diffusion approaches, analyzing their alignment mechanisms and innovations. At the task level, we illuminate new capabilities including conversational interaction, explainable reasoning, and personalized content generation. We identify five key advantages: world knowledge integration, natural language understanding, reasoning capabilities, scaling laws, and creative generation. We critically examine challenges in benchmark design, model robustness, and deployment efficiency, while charting a roadmap toward intelligent recommendation assistants that fundamentally reshape human-information interaction.
Large Language Models (LLMs) have transformed software development, enabling AI-powered applications known as LLM-based agents that promise to automate tasks across diverse apps and workflows. Yet, the security implications of deploying such agents in adversarial mobile environments remain poorly understood. In this paper, we present the first systematic study of security risks in mobile LLM agents. We design and evaluate a suite of adversarial case studies, ranging from opportunistic manipulations such as pop-up advertisements to advanced, end-to-end workflows involving malware installation and cross-app data exfiltration. Our evaluation covers eight state-of-the-art mobile agents across three architectures, with over 2,000 adversarial and paired benign trials. The results reveal systemic vulnerabilities: low-barrier vectors such as fraudulent ads succeed with over 80% reliability, while even workflows requiring the circumvention of operating-system warnings, such as malware installation, are consistently completed by advanced multi-app agents. By mapping these attacks to the MITRE ATT&CK Mobile framework, we uncover novel privilege-escalation and persistence pathways unique to LLM-driven automation. Collectively, our findings provide the first end-to-end evidence that mobile LLM agents are exploitable in realistic adversarial settings, where untrusted third-party channels (e.g., ads, embedded webviews, cross-app notifications) are an inherent part of the mobile ecosystem.
Artificial intelligence (AI) agents are emerging as transformative tools in drug discovery, with the ability to autonomously reason, act, and learn through complicated research workflows. Building on large language models (LLMs) coupled with perception, computation, action, and memory tools, these agentic AI systems could integrate diverse biomedical data, execute tasks, carry out experiments via robotic platforms, and iteratively refine hypotheses in closed loops. We provide a conceptual and technical overview of agentic AI architectures, ranging from ReAct and Reflection to Supervisor and Swarm systems, and illustrate their applications across key stages of drug discovery, including literature synthesis, toxicity prediction, automated protocol generation, small-molecule synthesis, drug repurposing, and end-to-end decision-making. To our knowledge, this represents the first comprehensive work to present real-world implementations and quantifiable impacts of agentic AI systems deployed in operational drug discovery settings. Early implementations demonstrate substantial gains in speed, reproducibility, and scalability, compressing workflows that once took months into hours while maintaining scientific traceability. We discuss the current challenges related to data heterogeneity, system reliability, privacy, and benchmarking, and outline future directions towards technology in support of science and translation.
With powerful and integrative large language models (LLMs), medical AI agents have demonstrated unique advantages in providing personalized medical consultations, continuous health monitoring, and precise treatment plans. Retrieval-Augmented Generation (RAG) integrates personal medical documents into LLMs by an external retrievable database to address the costly retraining or fine-tuning issues in deploying customized agents. While deploying medical agents in edge devices ensures privacy protection, RAG implementations impose substantial memory access and energy consumption during the retrieval stage. This paper presents a hierarchical retrieval architecture for edge RAG, leveraging a two-stage retrieval scheme that combines approximate retrieval for candidate set generation, followed by high-precision retrieval on pre-selected document embeddings. The proposed architecture significantly reduces energy consumption and external memory access while maintaining retrieval accuracy. Simulation results show that, under TSMC 28nm technology, the proposed hierarchical retrieval architecture has reduced the overall memory access by nearly 50% and the computation by 75% compared to pure INT8 retrieval, and the total energy consumption for 1 MB data retrieval is 177.76 {\mu}J/query.
State-of-the-art (SOTA) LLMs have progressed from struggling on proof-based Olympiad problems to solving most of the IMO 2025 problems, with leading systems reportedly handling 5 of 6 problems. Given this progress, we assess how well these models can grade proofs: detecting errors, judging their severity, and assigning fair scores beyond binary correctness. We study proof-analysis capabilities using a corpus of 90 Gemini 2.5 Pro-generated solutions that we grade on a 1-4 scale with detailed error annotations, and on MathArena solution sets for IMO/USAMO 2025 scored on a 0-7 scale. Our analysis shows that models can reliably flag incorrect (including subtly incorrect) solutions but exhibit calibration gaps in how partial credit is assigned. To address this, we introduce agentic workflows that extract and analyze reference solutions and automatically derive problem-specific rubrics for a multi-step grading process. We instantiate and compare different design choices for the grading workflows, and evaluate their trade-offs. Across our annotated corpus and MathArena, our proposed workflows achieve higher agreement with human grades and more consistent handling of partial credit across metrics. We release all code, data, and prompts/logs to facilitate future research.
With the increasing use of conversational AI systems, there is growing concern over privacy leaks, especially when users share sensitive personal data in interactions with Large Language Models (LLMs). Conversations shared with these models may contain Personally Identifiable Information (PII), which, if exposed, could lead to security breaches or identity theft. To address this challenge, we present the Local Optimizations for Pseudonymization with Semantic Integrity Directed Entity Detection (LOPSIDED) framework, a semantically-aware privacy agent designed to safeguard sensitive PII data when using remote LLMs. Unlike prior work that often degrade response quality, our approach dynamically replaces sensitive PII entities in user prompts with semantically consistent pseudonyms, preserving the contextual integrity of conversations. Once the model generates its response, the pseudonyms are automatically depseudonymized, ensuring the user receives an accurate, privacy-preserving output. We evaluate our approach using real-world conversations sourced from ShareGPT, which we further augment and annotate to assess whether named entities are contextually relevant to the model's response. Our results show that LOPSIDED reduces semantic utility errors by a factor of 5 compared to baseline techniques, all while enhancing privacy.
AI deployment increasingly resembles a pipeline of data transformation, fine-tuning, and agent interactions rather than a monolithic LLM job; recent examples include RLHF/RLAIF training and agentic workflows. To cope with this shift, we propose FlowMesh, a multi-tenant service fabric that executes and optimizes these workloads as one shared service instead of isolated pipelines. It decomposes workflows into fine-grained operators with recorded lineage, enabling de-duplication of work across users and batching requests on the same hardware while preserving per-workflow provenance. A global control plane maintains a cluster-wide pool of ready operators and uses a single utility function to pick both the batch and the worker, balancing throughput, cost, and data locality on heterogeneous GPUs. The data plane is an elastic fleet of stateless workers backed by a content-addressable store, enabling rapid, automatic scale-out, safe retry after preemption, and portability across managed clusters such as Kubernetes and geo-distributed GPU marketplaces such as Vast.ai. Compared with baseline solutions, FlowMesh achieves up to 3.8x cost reduction and 2.0x lower energy usage, provides a similar or better latency profile, and remains efficient under dynamic and failure-prone conditions.
As coding agents are increasingly deployed in large codebases, the need to automatically design challenging, codebase-level evaluation is central. We propose Gistify, a task where a coding LLM must create a single, minimal, self-contained file that can reproduce a specific functionality of a codebase. The coding LLM is given full access to a codebase along with a specific entrypoint (e.g., a python command), and the generated file must replicate the output of the same command ran under the full codebase, while containing only the essential components necessary to execute the provided command. Success on Gistify requires both structural understanding of the codebase, accurate modeling of its execution flow as well as the ability to produce potentially large code patches. Our findings show that current state-of-the-art models struggle to reliably solve Gistify tasks, especially ones with long executions traces.
Keeping software systems up to date is essential to avoid technical debt, security vulnerabilities, and the rigidity typical of legacy systems. However, updating libraries and frameworks remains a time consuming and error-prone process. Recent advances in Large Language Models (LLMs) and agentic coding systems offer new opportunities for automating such maintenance tasks. In this paper, we evaluate the update of a well-known Python library, SQLAlchemy, across a dataset of ten client applications. For this task, we use the Github's Copilot Agent Mode, an autonomous AI systema capable of planning and executing multi-step migration workflows. To assess the effectiveness of the automated migration, we also introduce Migration Coverage, a metric that quantifies the proportion of API usage points correctly migrated. The results of our study show that the LLM agent was capable of migrating functionalities and API usages between SQLAlchemy versions (migration coverage: 100%, median), but failed to maintain the application functionality, leading to a low test-pass rate (39.75%, median).
Multi-page visual documents such as manuals, brochures, presentations, and posters convey key information through layout, colors, icons, and cross-slide references. While large language models (LLMs) offer opportunities in document understanding, current systems struggle with complex, multi-page visual documents, particularly in fine-grained reasoning over elements and pages. We introduce SlideAgent, a versatile agentic framework for understanding multi-modal, multi-page, and multi-layout documents, especially slide decks. SlideAgent employs specialized agents and decomposes reasoning into three specialized levels-global, page, and element-to construct a structured, query-agnostic representation that captures both overarching themes and detailed visual or textual cues. During inference, SlideAgent selectively activates specialized agents for multi-level reasoning and integrates their outputs into coherent, context-aware answers. Extensive experiments show that SlideAgent achieves significant improvement over both proprietary (+7.9 overall) and open-source models (+9.8 overall).
Most scientific materials compress reasoning, presenting conclusions while omitting the derivational chains that justify them. This compression hinders verification by lacking explicit, step-wise justifications and inhibits cross-domain links by collapsing the very pathways that establish the logical and causal connections between concepts. We introduce a scalable framework that decompresses scientific reasoning, constructing a verifiable Long Chain-of-Thought (LCoT) knowledge base and projecting it into an emergent encyclopedia, SciencePedia. Our pipeline operationalizes an endpoint-driven, reductionist strategy: a Socratic agent, guided by a curriculum of around 200 courses, generates approximately 3 million first-principles questions. To ensure high fidelity, multiple independent solver models generate LCoTs, which are then rigorously filtered by prompt sanitization and cross-model answer consensus, retaining only those with verifiable endpoints. This verified corpus powers the Brainstorm Search Engine, which performs inverse knowledge search -- retrieving diverse, first-principles derivations that culminate in a target concept. This engine, in turn, feeds the Plato synthesizer, which narrates these verified chains into coherent articles. The initial SciencePedia comprises approximately 200,000 fine-grained entries spanning mathematics, physics, chemistry, biology, engineering, and computation. In evaluations across six disciplines, Plato-synthesized articles (conditioned on retrieved LCoTs) exhibit substantially higher knowledge-point density and significantly lower factual error rates than an equally-prompted baseline without retrieval (as judged by an external LLM). Built on this verifiable LCoT knowledge base, this reasoning-centric approach enables trustworthy, cross-domain scientific synthesis at scale and establishes the foundation for an ever-expanding encyclopedia.
The electricity sector transition requires substantial increases in residential demand response capacity, yet Home Energy Management Systems (HEMS) adoption remains limited by user interaction barriers requiring translation of everyday preferences into technical parameters. While large language models have been applied to energy systems as code generators and parameter extractors, no existing implementation deploys LLMs as autonomous coordinators managing the complete workflow from natural language input to multi-appliance scheduling. This paper presents an agentic AI HEMS where LLMs autonomously coordinate multi-appliance scheduling from natural language requests to device control, achieving optimal scheduling without example demonstrations. A hierarchical architecture combining one orchestrator with three specialist agents uses the ReAct pattern for iterative reasoning, enabling dynamic coordination without hardcoded workflows while integrating Google Calendar for context-aware deadline extraction. Evaluation across three open-source models using real Austrian day-ahead electricity prices reveals substantial capability differences. Llama-3.3-70B successfully coordinates all appliances across all scenarios to match cost-optimal benchmarks computed via mixed-integer linear programming, while other models achieve perfect single-appliance performance but struggle to coordinate all appliances simultaneously. Progressive prompt engineering experiments demonstrate that analytical query handling without explicit guidance remains unreliable despite models' general reasoning capabilities. We open-source the complete system including orchestration logic, agent prompts, tools, and web interfaces to enable reproducibility, extension, and future research.
Large Language Model (LLM) agents have evolved from basic text generation to autonomously completing complex tasks through interaction with external tools. However, current benchmarks mainly assess end-to-end performance in fixed scenarios, restricting evaluation to specific skills and suffering from score saturation and growing dependence on expert annotation as agent capabilities improve. In this work, we emphasize the importance of learning ability, including both self-improvement and peer-learning, as a core driver for agent evolution toward human-level intelligence. We propose an iterative, competitive peer-learning framework, which allows agents to refine and optimize their strategies through repeated interactions and feedback, thereby systematically evaluating their learning capabilities. To address the score saturation issue in current benchmarks, we introduce CATArena, a tournament-style evaluation platform featuring four diverse board and card games with open-ended scoring. By providing tasks without explicit upper score limits, CATArena enables continuous and dynamic evaluation of rapidly advancing agent capabilities. Experimental results and analyses involving both minimal and commercial code agents demonstrate that CATArena provides reliable, stable, and scalable benchmarking for core agent abilities, particularly learning ability and strategy coding.
While Multi-Agent Systems (MAS) excel at complex tasks, their growing autonomy with operational complexity often leads to critical inefficiencies, such as excessive token consumption and failures arising from misinformation. Existing methods primarily focus on post-hoc failure attribution, lacking proactive, real-time interventions to enhance robustness and efficiency. To this end, we introduce SupervisorAgent, a lightweight and modular framework for runtime, adaptive supervision that operates without altering the base agent's architecture. Triggered by an LLM-free adaptive filter, SupervisorAgent intervenes at critical junctures to proactively correct errors, guide inefficient behaviors, and purify observations. On the challenging GAIA benchmark, SupervisorAgent reduces the token consumption of the Smolagent framework by an average of 29.45% without compromising its success rate. Extensive experiments across five additional benchmarks (math reasoning, code generation, and question answering) and various SoTA foundation models validate the broad applicability and robustness of our approach. The code is available at https://github.com/LINs-lab/SupervisorAgent.