Large Language Models (LLMs) exhibit strong reasoning abilities for planning long-horizon, real-world tasks, yet existing agent benchmarks focus on task completion while neglecting time efficiency in parallel and asynchronous operations. To address this, we present ParaCook, a benchmark for time-efficient collaborative planning. Inspired by the Overcooked game, ParaCook provides an environment for various challenging interaction planning of multi-agent systems that are instantiated as cooking tasks, with a simplified action space to isolate the core challenge of strategic parallel planning. Through a comprehensive evaluation of state-of-the-art LLMs, we find that current approaches achieve suboptimal plans, which struggle with parallel actions or coordination. Our analysis also reveals LLMs' potential on abstract tasks where they can focus on high-level parallel optimization. ParaCook provides a scalable evaluation framework with adjustable complexity, establishing a foundation for developing and assessing time efficiency-aware multi-agent planning. The code and data are available at https://github.com/zsq259/ParaCook.
Recent advances in large language models (LLMs) have demonstrated remarkable capabilities in reasoning and tool utilization. However, the generalization of tool-augmented reinforcement learning (RL) across diverse domains remains underexplored. In this work, we investigate the cross-domain generalization of an LLM agent equipped with a code interpreter tool, which is exclusively trained on mathematical problem-solving tasks. Despite the restricted training domain, we evaluate the agent's performance across several distinct reasoning domains. The results reveal that RL-based tool usage learned from mathematical tasks can be effectively transferred to complex tasks in other domains, enabling great task performance and high token efficiency. To facilitate this cross-domain transfer, we propose a Tool Generalization Reinforcement Learning (TGRL) framework designed to promote domain-agnostic learning and skill migration, encompassing: (i) a standardized tool interface that abstracts domain-specific nuances through consistent formatting and explicit termination, fostering transferable invocation patterns; (ii) a dual-component reward system that decomposes rewards to incentivize generalizable behaviors like tool efficiency and reasoning abstraction, ensuring alignment and robustness across domain shifts; and (iii) an XML-based prompt template that separates thinking, tool calls, and responses to encourage modular, domain-invariant planning and coherent multi-turn interactions. Extensive experiments across diverse benchmarks validate our approach, achieving state-of-the-art performance and highlighting the cross-domain potential of Tool RL for LLM reasoning.
An agent facing a planning problem can use answers to how-to questions to reduce uncertainty and fill knowledge gaps, helping it solve both current and future tasks. However, their open ended nature, where valid answers to "How do I X?" range from executable actions to high-level descriptions of X's sub-goals, makes them challenging for AI agents to ask, and for AI experts to answer, in ways that support efficient planning. We introduce $How^{2}$, a memory agent framework that enables agents to ask how-to questions, store the answers, and reuse them for lifelong learning in interactive environments. We evaluate our approach in Plancraft, a Minecraft crafting environment, where agents must complete an assembly task by manipulating inventory items. Using teacher models that answer at varying levels of abstraction, from executable action sequences to high-level subgoal descriptions, we show that lifelong learning agents benefit most from answers that are abstracted and decoupled from the current state. $How^{2}$ offers a way for LLM-based agents to improve their planning capabilities over time by asking questions in interactive environments.
Multi-agent systems (MAS) and reinforcement learning (RL) are widely used to enhance the agentic capabilities of large language models (LLMs). MAS improves task performance through role-based orchestration, while RL uses environmental rewards to learn stronger policies, such as GRPO-style optimization. However, applying on-policy RL to MAS remains underexplored and presents unique challenges. Algorithmically, standard GRPO grouping assumptions break down because prompts vary by role and by turn. System-wise, the training stack must support MAS-workflow rollouts and on-policy updates for both single-policy and multi-policy models. We propose AT-GRPO, which includes (i) an agent- and turn-wise grouped RL algorithm tailored to MAS and (ii) a training system that supports both single- and multi-policy regimes. Across game, planning, coding, and math tasks, AT-GRPO delivers substantial gains. On long-horizon planning, it increases accuracy from a 14.0 to 47.0 percent single-agent RL baseline to 96.0 to 99.5 percent. It also improves reasoning performance, with average gains of 3.87 to 7.62 percent on coding tasks and 9.0 to 17.93 percent on math. Code and environments are available at: https://github.com/pettingllms-ai/PettingLLMs.
We introduce $\textbf{MASSE}$, the first Multi-Agent System for Structural Engineering, effectively integrating large language model (LLM)-based agents with real-world engineering workflows. Structural engineering is a fundamental yet traditionally stagnant domain, with core workflows remaining largely unchanged for decades despite its substantial economic impact and global market size. Recent advancements in LLMs have significantly enhanced their ability to perform complex reasoning, long-horizon planning, and precise tool utilization -- capabilities well aligned with structural engineering tasks such as interpreting design codes, executing load calculations, and verifying structural capacities. We present a proof-of-concept showing that most real-world structural engineering workflows can be fully automated through a training-free LLM-based multi-agent system. MASSE enables immediate deployment in professional environments, and our comprehensive validation on real-world case studies demonstrates that it can reduce expert workload from approximately two hours to mere minutes, while enhancing both reliability and accuracy in practical engineering scenarios.
Deep research frameworks have shown promising capabilities in synthesizing comprehensive reports from web sources. While deep research possesses significant potential to address complex issues through planning and research cycles, existing frameworks are deficient in sufficient evaluation procedures and stage-specific protections. They typically treat evaluation as exact match accuracy of question-answering, but overlook crucial aspects of report quality such as credibility, coherence, breadth, depth, and safety. This oversight may result in hazardous or malicious sources being integrated into the final report. To address these issues, we introduce DEEPRESEARCHGUARD, a comprehensive framework featuring four-stage safeguards with open-domain evaluation of references and reports. We assess performance across multiple metrics, e.g., defense success rate and over-refusal rate, and five key report dimensions. In the absence of a suitable safety benchmark, we introduce DRSAFEBENCH, a stage-wise benchmark for deep research safety. Our evaluation spans diverse state-of-the-art LLMs, including GPT-4o, Gemini-2.5-flash, DeepSeek-v3, and o4-mini. DEEPRESEARCHGUARD achieves an average defense success rate improvement of 18.16% while reducing over-refusal rate by 6%. The input guard provides the most substantial early-stage protection by filtering out obvious risks, while the plan and research guards enhance citation discipline and source credibility. Through extensive experiments, we show that DEEPRESEARCHGUARD enables comprehensive open-domain evaluation and stage-aware defenses that effectively block harmful content propagation, while systematically improving report quality without excessive over-refusal rates. The code can be found via https://github.com/Jasonya/DeepResearchGuard.
We introduce LLM x MapReduce-V3, a hierarchically modular agent system designed for long-form survey generation. Building on the prior work, LLM x MapReduce-V2, this version incorporates a multi-agent architecture where individual functional components, such as skeleton initialization, digest construction, and skeleton refinement, are implemented as independent model-context-protocol (MCP) servers. These atomic servers can be aggregated into higher-level servers, creating a hierarchically structured system. A high-level planner agent dynamically orchestrates the workflow by selecting appropriate modules based on their MCP tool descriptions and the execution history. This modular decomposition facilitates human-in-the-loop intervention, affording users greater control and customization over the research process. Through a multi-turn interaction, the system precisely captures the intended research perspectives to generate a comprehensive skeleton, which is then developed into an in-depth survey. Human evaluations demonstrate that our system surpasses representative baselines in both content depth and length, highlighting the strength of MCP-based modular planning.
Robots navigating dynamic, cluttered, and semantically complex environments must integrate perception, symbolic reasoning, and spatial planning to generalize across diverse layouts and object categories. Existing methods often rely on static priors or limited memory, constraining adaptability under partial observability and semantic ambiguity. We present GRIP, Grid-based Relay with Intermediate Planning, a unified, modular framework with three scalable variants: GRIP-L (Lightweight), optimized for symbolic navigation via semantic occupancy grids; GRIP-F (Full), supporting multi-hop anchor chaining and LLM-based introspection; and GRIP-R (Real-World), enabling physical robot deployment under perceptual uncertainty. GRIP integrates dynamic 2D grid construction, open-vocabulary object grounding, co-occurrence-aware symbolic planning, and hybrid policy execution using behavioral cloning, D* search, and grid-conditioned control. Empirical results on AI2-THOR and RoboTHOR benchmarks show that GRIP achieves up to 9.6% higher success rates and over $2\times$ improvement in path efficiency (SPL and SAE) on long-horizon tasks. Qualitative analyses reveal interpretable symbolic plans in ambiguous scenes. Real-world deployment on a Jetbot further validates GRIP's generalization under sensor noise and environmental variation. These results position GRIP as a robust, scalable, and explainable framework bridging simulation and real-world navigation.
We present an approach to software testing automation using Agentic Retrieval-Augmented Generation (RAG) systems for Quality Engineering (QE) artifact creation. We combine autonomous AI agents with hybrid vector-graph knowledge systems to automate test plan, case, and QE metric generation. Our approach addresses traditional software testing limitations by leveraging LLMs such as Gemini and Mistral, multi-agent orchestration, and enhanced contextualization. The system achieves remarkable accuracy improvements from 65% to 94.8% while ensuring comprehensive document traceability throughout the quality engineering lifecycle. Experimental validation of enterprise Corporate Systems Engineering and SAP migration projects demonstrates an 85% reduction in testing timeline, an 85% improvement in test suite efficiency, and projected 35% cost savings, resulting in a 2-month acceleration of go-live.
Large Language Models (LLMs) are increasingly deployed in mental health contexts, from structured therapeutic support tools to informal chat-based well-being assistants. While these systems increase accessibility, scalability, and personalization, their integration into mental health care brings privacy and safety challenges that have not been well-examined. Unlike traditional clinical interactions, LLM-mediated therapy often lacks a clear structure for what information is collected, how it is processed, and how it is stored or reused. Users without clinical guidance may over-disclose personal information, which is sometimes irrelevant to their presenting concern, due to misplaced trust, lack of awareness of data risks, or the conversational design of the system. This overexposure raises privacy concerns and also increases the potential for LLM bias, misinterpretation, and long-term data misuse. We propose a framework embedding Artificial Intelligence (AI) literacy interventions directly into mental health conversational systems, and outline a study plan to evaluate their impact on disclosure safety, trust, and user experience.
Large Language Models (LLMs) can help robots reason about abstract task specifications. This requires augmenting classical representations of the environment used by robots with natural language-based priors. There are a number of existing approaches to doing so, but they are tailored to specific tasks, e.g., visual-language models for navigation, language-guided neural radiance fields for mapping, etc. This paper argues that the Universal Scene Description (USD) language is an effective and general representation of geometric, photometric and semantic information in the environment for LLM-based robotics tasks. Our argument is simple: a USD is an XML-based scene graph, readable by LLMs and humans alike, and rich enough to support essentially any task -- Pixar developed this language to store assets, scenes and even movies. We demonstrate a ``Real to USD'' system using a Unitree Go2 quadruped robot carrying LiDAR and a RGB camera that (i) builds an explicit USD representation of indoor environments with diverse objects and challenging settings with lots of glass, and (ii) parses the USD using Google's Gemini to demonstrate scene understanding, complex inferences, and planning. We also study different aspects of this system in simulated warehouse and hospital settings using Nvidia's Issac Sim. Code is available at https://github.com/grasp-lyrl/Real2USD .
Motion planning in complex scenarios is a core challenge in autonomous driving. Conventional methods apply predefined rules or learn from driving data to generate trajectories, while recent approaches leverage large language models (LLMs) for decision-making. However, it remains unclear whether LLMs truly capture human driving logic. We propose Align2Act, a motion planning framework that transforms instruction-tuned LLMs into interpretable planners aligned with human behavior. We derive structured driving instructions based on human reasoning patterns (e.g., anticipate hazards, yield at intersections) and traffic rules (e.g., stop at red lights, maintain lane boundaries). Our Align2ActChain module guides step-by-step reasoning to produce both an interpretable rationale and a safe trajectory. By fine-tuning LLaMA-2-7B with LoRA on one million scenarios from the nuPlan dataset, our method achieves an open-loop score of 85.17 and closed-loop scores of 70.31 (non-reactive) and 66.96 (reactive) on Test14-random. Unlike prior work focused on synthetic or open-loop settings, we demonstrate improved planning quality and human-likeness on the real-world nuPlan closed-loop benchmark. Ablation studies confirm that structured reasoning significantly improves performance over baseline LLM planners.
Multi-agent systems (MASs) have emerged as a promising paradigm for automated code generation, demonstrating impressive performance on established benchmarks by decomposing complex coding tasks across specialized agents with different roles. Despite their prosperous development and adoption, their robustness remains pressingly under-explored, raising critical concerns for real-world deployment. This paper presents the first comprehensive study examining the robustness of MASs for code generation through a fuzzing-based testing approach. By designing a fuzzing pipeline incorporating semantic-preserving mutation operators and a novel fitness function, we assess mainstream MASs across multiple datasets and LLMs. Our findings reveal substantial robustness flaws of various popular MASs: they fail to solve 7.9%-83.3% of problems they initially resolved successfully after applying the semantic-preserving mutations. Through comprehensive failure analysis, we identify a common yet largely overlooked cause of the robustness issue: miscommunications between planning and coding agents, where plans lack sufficient detail and coding agents misinterpret intricate logic, aligning with the challenges inherent in a multi-stage information transformation process. Accordingly, we also propose a repairing method that encompasses multi-prompt generation and introduces a new monitor agent to address this issue. Evaluation shows that our repairing method effectively enhances the robustness of MASs by solving 40.0%-88.9% of identified failures. Our work uncovers critical robustness flaws in MASs and provides effective mitigation strategies, contributing essential insights for developing more reliable MASs for code generation.
Coordinating heterogeneous robot fleets to achieve multiple goals is challenging in multi-robot systems. We introduce an open-source and extensible framework for centralized multi-robot task planning and scheduling that leverages LLMs to enable fleets of heterogeneous robots to accomplish multiple tasks. RobotFleet provides abstractions for planning, scheduling, and execution across robots deployed as containerized services to simplify fleet scaling and management. The framework maintains a shared declarative world state and two-way communication for task execution and replanning. By modularizing each layer of the autonomy stack and using LLMs for open-world reasoning, RobotFleet lowers the barrier to building scalable multi-robot systems. The code can be found here: https://github.com/therohangupta/robot-fleet.
The rapid advancement of large language models (LLMs) has revolutionized artificial intelligence, shifting from supporting objective tasks (e.g., recognition) to empowering subjective decision-making (e.g., planning, decision). This marks the dawn of general and powerful AI, with applications spanning a wide range of fields, including programming, education, healthcare, finance, and law. However, their deployment introduces multifaceted risks. Due to the black-box nature of LLMs and the human-like quality of their generated content, issues such as hallucinations, bias, unfairness, and copyright infringement become particularly significant. In this context, sourcing information from multiple perspectives is essential. This survey presents a systematic investigation into provenance tracking for content generated by LLMs, organized around four interrelated dimensions that together capture both model- and data-centric perspectives. From the model perspective, Model Sourcing treats the model as a whole, aiming to distinguish content generated by specific LLMs from content authored by humans. Model Structure Sourcing delves into the internal generative mechanisms, analyzing architectural components that shape the outputs of model. From the data perspective, Training Data Sourcing focuses on internal attribution, tracing the origins of generated content back to the training data of model. In contrast, External Data Sourcing emphasizes external validation, identifying external information used to support or influence the responses of model. Moreover, we also propose a dual-paradigm taxonomy that classifies existing sourcing methods into prior-based (proactive traceability embedding) and posterior-based (retrospective inference) approaches. Traceability across these dimensions enhances the transparency, accountability, and trustworthiness of LLMs deployment in real-world applications.
Large language models (LLMs) enable end-users to delegate complex tasks to autonomous agents through natural language. However, prompt-based interaction faces critical limitations: Users often struggle to specify procedural requirements for tasks, especially those that don't have a factually correct solution but instead rely on personal preferences, such as posting social media content or planning a trip. Additionally, a ''successful'' prompt for one task may not be reusable or generalizable across similar tasks. We present ALLOY, a system inspired by classical HCI theories on Programming by Demonstration (PBD), but extended to enhance adaptability in creating LLM-based web agents. ALLOY enables users to express procedural preferences through natural demonstrations rather than prompts, while making these procedures transparent and editable through visualized workflows that can be generalized across task variations. In a study with 12 participants, ALLOY's demonstration--based approach outperformed prompt-based agents and manual workflows in capturing user intent and procedural preferences in complex web tasks. Insights from the study also show how demonstration--based interaction complements the traditional prompt-based approach.
LLM-powered agents are both a promising new technology and a source of complexity, where choices about models, tools, and prompting can affect their usefulness. While numerous benchmarks measure agent accuracy across domains, they mostly assume full automation, failing to represent the collaborative nature of real-world use cases. In this paper, we make two major steps towards the rigorous assessment of human-agent interactions. First, we propose PULSE, a framework for more efficient human-centric evaluation of agent designs, which comprises collecting user feedback, training an ML model to predict user satisfaction, and computing results by combining human satisfaction ratings with model-generated pseudo-labels. Second, we deploy the framework on a large-scale web platform built around the open-source software agent OpenHands, collecting in-the-wild usage data across over 15k users. We conduct case studies around how three agent design decisions -- choice of LLM backbone, planning strategy, and memory mechanisms -- impact developer satisfaction rates, yielding practical insights for software agent design. We also show how our framework can lead to more robust conclusions about agent design, reducing confidence intervals by 40\% compared to a standard A/B test. Finally, we find substantial discrepancies between in-the-wild results and benchmark performance (e.g., the anti-correlation between results comparing claude-sonnet-4 and gpt-5), underscoring the limitations of benchmark-driven evaluation. Our findings provide guidance for evaluations of LLM agents with humans and identify opportunities for better agent designs.
While LLM agents can plan multi-step tasks, intervening at the planning stage-before any action is executed-is often the safest way to prevent harm, since certain risks can lead to severe consequences once carried out. However, existing guardrails mostly operate post-execution, which is difficult to scale and leaves little room for controllable supervision at the plan level. To address this challenge, we highlight three critical gaps in current research: data gap, model gap, and evaluation gap. To close the data gap, we introduce AuraGen, a controllable engine that (i) synthesizes benign trajectories, (ii) injects category-labeled risks with calibrated difficulty, and (iii) filters outputs via an automated reward model, producing large and reliable corpora for pre-execution safety. To close the guardian model gap, we propose a foundational guardrail Safiron, combining a cross-planner adapter with a compact guardian model. The adapter unifies different input formats, while Safiron flags risky cases, assigns risk types, and generates rationales; trained in two stages with a broadly explored data recipe, Safiron achieves robust transfer across settings. To close the evaluation gap, we release Pre-Exec Bench, a realistic benchmark covering diverse tools and branching trajectories, which measures detection, fine-grained categorization, explanation, and cross-planner generalization in human-verified scenarios. Extensive experiments demonstrate consistent gains of the proposed guardrail over strong baselines on Pre-Exec Bench, and ablations further distill actionable practices, providing a practical template for safer agentic systems.
Competitive programming problems increasingly serve as valuable benchmarks to evaluate the coding capabilities of large language models (LLMs) due to their complexity and ease of verification. Yet, current coding benchmarks face limitations such as lack of exceptionally challenging problems, insufficient test case coverage, reliance on online platform APIs that limit accessibility. To address these issues, we introduce LiveOIBench, a comprehensive benchmark featuring 403 expert-curated Olympiad-level competitive programming problems, each with an average of 60 expert-designed test cases. The problems are sourced directly from 72 official Informatics Olympiads in different regions conducted between 2023 and 2025. LiveOIBench distinguishes itself through four key features: (1) meticulously curated high-quality tasks with detailed subtask rubrics and extensive private test cases; (2) direct integration of elite contestant performance data to enable informative comparison against top-performing humans; (3) planned continuous, contamination-free updates from newly released Olympiad problems; and (4) a self-contained evaluation system facilitating offline and easy-to-reproduce assessments. Benchmarking 32 popular general-purpose and reasoning LLMs, we find that GPT-5 achieves a notable 81.76th percentile, a strong result that nonetheless falls short of top human contestant performance, who usually place above 90th. In contrast, among open-weight reasoning models, GPT-OSS-120B achieves only a 60th percentile, underscoring significant capability disparities from frontier closed models. Detailed analyses indicate that robust reasoning models prioritize precise problem analysis over excessive exploration, suggesting future models should emphasize structured analysis and minimize unnecessary exploration. All data, code, and leaderboard results will be made publicly available on our website.
Building agents, systems that perceive and act upon their environment with a degree of autonomy, has long been a focus of AI research. This pursuit has recently become vastly more practical with the emergence of large language models (LLMs) capable of using natural language to integrate information, follow instructions, and perform forms of "reasoning" and planning across a wide range of tasks. With its multimodal data streams and orchestrated workflows spanning multiple systems, radiology is uniquely suited to benefit from agents that can adapt to context and automate repetitive yet complex tasks. In radiology, LLMs and their multimodal variants have already demonstrated promising performance for individual tasks such as information extraction and report summarization. However, using LLMs in isolation underutilizes their potential to support complex, multi-step workflows where decisions depend on evolving context from multiple information sources. Equipping LLMs with external tools and feedback mechanisms enables them to drive systems that exhibit a spectrum of autonomy, ranging from semi-automated workflows to more adaptive agents capable of managing complex processes. This review examines the design of such LLM-driven agentic systems, highlights key applications, discusses evaluation methods for planning and tool use, and outlines challenges such as error cascades, tool-use efficiency, and health IT integration.
This paper reviews the architecture and implementation methods of agents powered by large language models (LLMs). Motivated by the limitations of traditional LLMs in real-world tasks, the research aims to explore patterns to develop "agentic" LLMs that can automate complex tasks and bridge the performance gap with human capabilities. Key components include a perception system that converts environmental percepts into meaningful representations; a reasoning system that formulates plans, adapts to feedback, and evaluates actions through different techniques like Chain-of-Thought and Tree-of-Thought; a memory system that retains knowledge through both short-term and long-term mechanisms; and an execution system that translates internal decisions into concrete actions. This paper shows how integrating these systems leads to more capable and generalized software bots that mimic human cognitive processes for autonomous and intelligent behavior.
The integration of LLMs into software engineering has catalyzed a paradigm shift from traditional rule-based systems to sophisticated agentic systems capable of autonomous problem-solving. Despite this transformation, the field lacks a comprehensive understanding of how benchmarks and solutions interconnect, hindering systematic progress and evaluation. This survey presents the first holistic analysis of LLM-empowered software engineering, bridging the critical gap between evaluation and solution approaches. We analyze 150+ recent papers and organize them into a comprehensive taxonomy spanning two major dimensions: (1) Solutions, categorized into prompt-based, fine-tuning-based, and agent-based paradigms, and (2) Benchmarks, covering code generation, translation, repair, and other tasks. Our analysis reveals how the field has evolved from simple prompt engineering to complex agentic systems incorporating planning and decomposition, reasoning and self-refinement, memory mechanisms, and tool augmentation. We present a unified pipeline that illustrates the complete workflow from task specification to final deliverables, demonstrating how different solution paradigms address varying complexity levels across software engineering tasks. Unlike existing surveys that focus on isolated aspects, we provide full-spectrum coverage connecting 50+ benchmarks with their corresponding solution strategies, enabling researchers to identify optimal approaches for specific evaluation criteria. Furthermore, we identify critical research gaps and propose actionable future directions, including multi-agent collaboration frameworks, self-evolving code generation systems, and integration of formal verification with LLM-based methods. This survey serves as a foundational resource for researchers and practitioners seeking to understand, evaluate, and advance LLM-empowered software engineering systems.
Travel planning is a valuable yet complex task that poses significant challenges even for advanced large language models (LLMs). While recent benchmarks have advanced in evaluating LLMs' planning capabilities, they often fall short in evaluating feasibility, reliability, and engagement of travel plans. We introduce a comprehensive benchmark for travel planning that unifies fine-grained criteria into a single reward, enabling direct comparison of plan quality and seamless integration with reinforcement learning (RL). Our evaluator achieves moderate agreement with travel-expert annotations (60.75\%) and outperforms multiple LLM-as-judge baselines. We further release a large-scale dataset of 4,870 queries including 219 real-world, free-form requests for generalization to authentic user intent. Using this benchmark, we conduct extensive experiments across diverse methods and LLMs, including test-time computation, neuro-symbolic approaches, supervised fine-tuning, and RL via GRPO. Across base models, RL generally improves itinerary feasibility over prompt-only and supervised baselines, yielding higher unified reward scores.
While researchers have made significant progress in enabling large language models (LLMs) to perform multi-step planning, LLMs struggle to ensure that those plans align with high-level user intent and satisfy symbolic constraints, especially in complex, multi-step domains. Existing reasoning approaches such as Chain-of-Thought (CoT), Tree-of-Thought (ToT), and verifier-augmented methods, expand the search space but often yield infeasible actions or hallucinated steps. To overcome these limitations, we propose Constraints-of-Thought (Const-o-T), a framework that provides a structured prior that enables Monte Carlo Tree Search (MCTS) focus search on semantically meaningful paths. Each reasoning step is represented as an (intent, constraint) pair, which serves both to compress the search space and enforce validity. Unlike prior methods that merely generate reasoning traces or validate outputs post hoc, Const-o-T uses (intent, constraint)pairs to actively focus the search toward feasible and meaningful plans. We integrate Const-o-T into MCTS using a structured representation of intent-constraint pairs constraints prune infeasible branches and guide exploration toward semantically valid actions, improving planning efficiency and verifiable decision-making. We demonstrate across three domains Risk game, CAD code generation, and arithmetic reasoning that our approach outperforms baselines, yielding higher accuracy and stronger structural alignment. Our contribution is to demonstrate that Const-of-T offers a generalizable foundation for constraint-guided reasoning, enabling more efficient, constraint-aligned, and domain-adaptable planning with LLMs.
Text-attributed graphs (TAGs) have emerged as a powerful representation that combines structural connections with fine-grained semantics, supporting a wide range of data-centric applications. However, the performance of graph neural networks (GNNs) on TAGs is highly sensitive to input quality. Our empirical study shows that both traditional GNNs and LLM-enhanced GNNs suffer significant degradation across nine representative scenarios of sparsity, noise, and imbalance, highlighting graph quality as a critical bottleneck. Existing approaches mainly focus on improving model architectures, while neglecting systematic optimization of TAG data itself, leading to limited effectiveness in practice. To address this gap, we propose LAGA (Large Language and Graph Agent), a unified multi-agent framework that treats graph quality control as a first-class, data-centric problem. LAGA integrates four collaborative agents-detection, planning, action, and evaluation-into an automated closed loop. At its core, the action agent employs a dual-encoder and tri-objective design to capture complementary information across modalities and perform holistic graph quality enhancement. Experiments across nine scenarios show that LAGA improves graph quality and achieves state-of-the-art performance across various tasks and backbones, validating data-centric quality optimization as key to reliable TAGs and robust graph learning.
We introduce the Agent GPA (Goal-Plan-Action) framework: an evaluation paradigm based on an agent's operational loop of setting goals, devising plans, and executing actions. The framework includes five evaluation metrics: Goal Fulfillment, Logical Consistency, Execution Efficiency, Plan Quality, and Plan Adherence. Logical Consistency checks that an agent's actions are consistent with its prior actions. Execution Efficiency checks whether the agent executes in the most efficient way to achieve its goal. Plan Quality checks whether an agent's plans are aligned with its goals; Plan Adherence checks if an agent's actions are aligned with its plan; and Goal Fulfillment checks that agent's final outcomes match the stated goals. Our experimental results on two benchmark datasets - the public TRAIL/GAIA dataset and an internal dataset for a production-grade data agent - show that this framework (a) provides a systematic way to cover a broad range of agent failures, including all agent errors on the TRAIL/GAIA benchmark dataset; (b) supports LLM-judges that exhibit strong agreement with human annotation, covering 80% to over 95% errors; and (c) localizes errors with 86% agreement to enable targeted improvement of agent performance.
Large language models (LLMs) have demonstrated impressive reasoning abilities yet remain unreliable on knowledge-intensive, multi-hop questions -- they miss long-tail facts, hallucinate when uncertain, and their internal knowledge lags behind real-world change. Knowledge graphs (KGs) offer a structured source of relational evidence, but existing KGQA methods face fundamental trade-offs: compiling complete SPARQL queries without knowing available relations proves brittle, retrieving large subgraphs introduces noise, and complex agent frameworks with parallel exploration exponentially expand search spaces. To address these limitations, we propose Search-on-Graph (SoG), a simple yet effective framework that enables LLMs to perform iterative informed graph navigation using a single, carefully designed \textsc{Search} function. Rather than pre-planning paths or retrieving large subgraphs, SoG follows an ``observe-then-navigate'' principle: at each step, the LLM examines actual available relations from the current entity before deciding on the next hop. This approach further adapts seamlessly to different KG schemas and handles high-degree nodes through adaptive filtering. Across six KGQA benchmarks spanning Freebase and Wikidata, SoG achieves state-of-the-art performance without fine-tuning. We demonstrate particularly strong gains on Wikidata benchmarks (+16\% improvement over previous best methods) alongside consistent improvements on Freebase benchmarks.
Human-robot collaboration (HRC) requires robots to adapt their motions to human intent to ensure safe and efficient cooperation in shared spaces. Although large language models (LLMs) provide high-level reasoning for inferring human intent, their application to reliable motion planning in HRC remains challenging. Physical human-robot interaction (pHRI) is intuitive but often relies on continuous kinesthetic guidance, which imposes burdens on operators. To address these challenges, a contact-informed adaptive motion-planning framework is introduced to infer human intent directly from physical contact and employ the inferred intent for online motion correction in HRC. First, an optimization-based force estimation method is proposed to infer human-intended contact forces and locations from joint torque measurements and a robot dynamics model, thereby reducing cost and installation complexity while enabling whole-body sensitivity. Then, a torque-based contact detection mechanism with link-level localization is introduced to reduce the optimization search space and to enable real-time estimation. Subsequently, a contact-informed adaptive motion planner is developed to infer human intent from contacts and to replan robot motion online, while maintaining smoothness and adapting to human corrections. Finally, experiments on a 7-DOF manipulator are conducted to demonstrate the accuracy of the proposed force estimation method and the effectiveness of the contact-informed adaptive motion planner under perception uncertainty in HRC.
Long-horizon tasks that require sustained reasoning and multiple tool interactions remain challenging for LLM agents: small errors compound across steps, and even state-of-the-art models often hallucinate or lose coherence. We identify context management as the central bottleneck -- extended histories cause agents to overlook critical evidence or become distracted by irrelevant information, thus failing to replan or reflect from previous mistakes. To address this, we propose COMPASS (Context-Organized Multi-Agent Planning and Strategy System), a lightweight hierarchical framework that separates tactical execution, strategic oversight, and context organization into three specialized components: (1) a Main Agent that performs reasoning and tool use, (2) a Meta-Thinker that monitors progress and issues strategic interventions, and (3) a Context Manager that maintains concise, relevant progress briefs for different reasoning stages. Across three challenging benchmarks -- GAIA, BrowseComp, and Humanity's Last Exam -- COMPASS improves accuracy by up to 20% relative to both single- and multi-agent baselines. We further introduce a test-time scaling extension that elevates performance to match established DeepResearch agents, and a post-training pipeline that delegates context management to smaller models for enhanced efficiency.
Reinforcement Learning (RL) agents often struggle in sparse-reward environments where traditional exploration strategies fail to discover effective action sequences. Large Language Models (LLMs) possess procedural knowledge and reasoning capabilities from text pretraining that could guide RL exploration, but existing approaches create rigid dependencies where RL policies must follow LLM suggestions or incorporate them directly into reward functions. We propose a framework that provides LLM-generated action recommendations through augmented observation spaces, allowing RL agents to learn when to follow or ignore this guidance. Our method leverages LLMs' world knowledge and reasoning abilities while maintaining flexibility through soft constraints. We evaluate our approach on three BabyAI environments of increasing complexity and show that the benefits of LLM guidance scale with task difficulty. In the most challenging environment, we achieve 71% relative improvement in final success rates over baseline. The approach provides substantial sample efficiency gains, with agents reaching performance thresholds up to 9 times faster, and requires no modifications to existing RL algorithms. Our results demonstrate an effective method for leveraging LLM planning capabilities to accelerate RL training in challenging environments.