Search research

This paper demonstrates how deep reinforcement learning (DRL) enables adaptive locomotion of snake-like robots in dynamically changing viscous environments, overcoming the inherent performance limitations of classical predefined control methods. The lack of direct onboard sensors for fluid properties necessitates formulating this task as a partially observable Markov decision process. By employing an asymmetric actor-critic framework, a teacher policy trained using privileged information available only in the physics simulator distills its knowledge into a student policy that relies solely on proprioceptive sensor information. Simulation results across a wide range of dynamic viscosity changes ($10^{-7}$ to $10^{-2} m^2/s$) reveal that the DRL agent autonomously acquires non-sinusoidal adaptive gaits. These gaits improve propulsion velocity and transport efficiency, breaking the inherent limits of conventional sinusoidal and kinematic control. The findings establish that implicit environment inference via privileged information distillation is an effective approach to bypass the constraints of classical models under unpredictable fluid dynamics.

Navigating dynamic and crowded environments presents significant challenges for quadruped robots due to severe sensor occlusion and unpredictable human motion. Existing approaches face a trade-off: model-based methods, such as Velocity Obstacles (VO), theoretically guarantee safety but rely on accurate obstacle motion estimates that often fail in dense crowds, while end-to-end learning methods offer robustness but lack motion prediction capability of obstacles, leading to collisions or conservative behaviors. To solve this, we propose VOP-Nav, a novel navigation system that combines the geometric safety of VO with the agile adaptability of end-to-end learning. Using only local onboard observations, our system avoids explicit obstacle detection and tracking pipelines. The VOP-Net processes multi-frame LiDAR data to implicitly encode dynamic constraints and predict a safe velocity region derived from Velocity Obstacle theory. Importantly, the VO predictions serve a dual role: they are used as input to the navigation policy during inference and as a reward signal during training to encourage safe motion. Evaluations in Isaac Gym demonstrate that VOP-Nav achieves higher success rates than all baselines while balancing locomotion speed and collision avoidance. Real-world deployment on a Unitree Go2 quadruped robot further validates the system's robustness and efficiency in complex indoor and outdoor dynamic environments.

This paper presents a cascaded Low-Rank Adaptation (LoRA)-based multimodal fusion framework for action and activity recognition in healthcare-oriented training environments. The proposed architecture combines parameter-efficient modality-specific adaptation with sequential fusion, enabling modalities to be integrated in stages without retraining previously learned components. Rather than assuming a fixed fusion structure, the framework first integrates more closely related modalities and then incorporates additional heterogeneous modalities, supporting scalable adaptation across datasets with different modality sets.We evaluate the framework on two healthcare-oriented training environment datasets: NurViD and the Nurse Training dataset. Across these datasets, preliminary results suggest that the proposed cascaded fusion strategy improves over individual modality models and provides competitive performance relative to previously reported dataset-specific baselines. Overall, these findings indicate that cascaded LoRA-based fusion is a promising parameter-efficient approach for integrating heterogeneous modalities in medical training action and activity recognition tasks. github: https://github.com/anonymous0-ai/LoRA-Based-Cascaded-Multimodal-Fusion-.git.

Multi-object search in unknown household environments requires planning under extensive uncertainty - from unknown object locations to cluttered spaces with unobserved obstacles. POMDPs offer a principled framework for such problems but remain intractable in large domains. We propose Inter-POMDP, a novel interleaved POMDP planning algorithm that decomposes this challenge into two interacting levels: a high-level POUCT planner reasons over object distributions using LLM-informed histogram beliefs, while a low-level motion planner models navigation uncertainty with obstacle-aware particle beliefs as domain knowledge to guide high-level POUCT. This interleaved design balances planning quality and efficiency despite the large search space across unknown multi-room environments. Both simulation and real-world experiments show that our Inter-POMDP algorithm reduces collision counts by up to 63%, navigation steps by up to 35%, and detection counts by up to 32% compared with baseline methods. Full videos are https://sites.google.com/view/inter-pomdp

Multi-robot path planning in human-shared environments requires a delicate balance between robust inter-robot coordination and socially aware behavior. While diffusion models excel at generating predictable, human-like paths, existing generative planners are often restricted to paths of fixed duration and high computational latency, limiting their adaptability to varying goal distances and hindering real-time deployment. We present Multi-Robot Rolling Diffusion (MRRD), a novel framework that enables real-time, long-horizon navigation for large robot teams through dense crowds. MRRD combines a rolling-horizon scheme to accommodate the limited prediction horizon of human motion, parallelized diffusion inference for scalable generation of human-like paths, and a conflict-based-search mechanism for resolving inter-robot collisions. It further incorporates urgency-based temporal conditioning to generate paths with varying speeds and employs differentiated guidance terms to maximize both social awareness around humans and efficient coordination between robots. Experimental results in crowded environments demonstrate that MRRD successfully scales to 15 robots in real-time, significantly outperforming existing baselines in both safety and mission success rates.

Autonomous agents promise substantial gains in speed, scale, and labor efficiency, but their failures can impose abrupt and often irreversible costs. The central challenge for agentic automation is therefore to increase the benefits of automation while bounding the consequences of failure. While databases remain central to modern computing, agents operate over a broader data environment spanning files, APIs, applications, and system state. In this talk, I will outline early work on Agentic Data Environments -- the execution substrate in which agents operate -- that both amplify agent capabilities and enforce safety guarantees. This perspective reframes data systems from passive stores of state into active substrates for safe, reliable execution.

Dynamic environments remain a fundamental challenge for visual SLAM, where unreliable observations from moving objects and rapid motion degrade state estimation accuracy. Although event cameras preserve fine-grained spatio-temporal information, most existing event-based SLAM frameworks still assume static scenes and lack approaches to estimate the reliability of features. To this end, we propose PLED-VINS, a monocular event camera-based visual-inertial SLAM framework that enables robust state estimation in dynamic environments. We propose an entropy-recency score map to characterize the temporal reliability of both point and line features based on event temporal statistics. Concurrently, geometric reliability is estimated via a unified point-line robust bundle adjustment. Building upon these, we design an adaptive weighting strategy that fuses temporal and geometric reliability, including motion-conditioned reliability modeling for line features, to suppress unreliable observations. Experimental results demonstrate that PLED-VINS improves state estimation on the evaluated dynamic sequences with moving objects.

Robotic throwing enables fast and efficient object placement beyond the robot's immediate workspace, but reliable throwing in cluttered environments remains underexplored. Existing approaches, such as TossingBot, learn throwing strategies from visual input but assume obstacle-free settings. In this paper, we address the problem of throwing objects into a target basket while avoiding obstacles placed randomly in the scene. We introduce a potential field state representation that compactly encodes both basket attraction and obstacle repulsion on a fixed-size grid, enabling reinforcement learning (RL) policies to generalize across arbitrary numbers and configurations of obstacles. The policy is initialized from kinesthetic demonstrations and optimized in simulation using three state-of-the-art RL algorithms (SAC, DDPG, TD3). Among these, SAC achieves the most consistent performance across scenarios. We compare the potential field representation against explicit state encodings and demonstrate that it achieves higher success rates and better scalability to unseen obstacle configurations. Real-robot experiments with unseen throwable objects confirm robust sim-to-real transfer, achieving up to $90\%$ success in cluttered scenes. These results demonstrate that PFR provides a practical and robust representation for safe and efficient robotic throwing in unstructured environments. A video showcasing our experiments is available at: https://youtu.be/ZZnJf8ua2dE

As Large Language Models (LLMs) evolve into autonomous agents, traditional static evaluation fails to capture multi-step decision-making. We introduce AgenticAI-Supervisor, an API and UI-driven RL Gym environment that decouples environment creation from scalable execution. By moving to verifiable execution outcomes, the platform generates high-fidelity traces and applies multi-dimensional reward shaping. Critically, our framework mitigates reward hacking through rigorous internal state validation and testing. This work provides a first look at our platform's core capabilities through a Customer Support Agent case study demonstrating a consistent closed-loop feedback for model optimization. Future work will focus on advanced features such as Computer Use, Tool Use, automated "stumping", and edge-case generation.

Language agents, i.e., LLM agents, progress rapidly and are increasingly deployed in production environments. This trend underscores the urgent need for rigorous and realistic evaluations. However, most existing benchmarks evaluate agents in simplified, idealized settings. They typically rely on pre-packaged tool interfaces, overlook critical steps, and assume inputs are clean and fully specified. Consequently, they understate the difficulty of real deployments, where uncertainty and noise are ubiquitous and agents must proactively explore the environment to uncover new tools. To bridge this gap, we present AgentGym2, a new evaluation framework with task instances grounded in real-world end-to-end working demands. Beyond reasoning and planning, it measures agents' ability to execute end-to-end procedures, discover tools via exploration, compose tools for unseen tasks, and remain robust to noisy and underspecified information. Experiments on 15 proprietary and open-source models show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2, revealing a substantial gap between the capability of current agents and the demands of real-world applications.

Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress. We introduce Autonomous Policy Evolution, a controlled evaluation setting in which a harness-model agent repeatedly edits an executable policy system under a fixed interaction budget. We instantiate this setting in EvoPolicyGym, a benchmark built from compact interactive RL environments that evaluates how agents iteratively improve explored policies. On the EvoPolicyGym suite, GPT-5.5 achieves the strongest aggregate rank score and top-two performance on all 16 environments. Beyond leaderboard results, EvoPolicyGym also provides trajectory-level diagnostics that distinguish how agents allocate budget, convert feedback into parametric tuning. These analyses show that strong autonomous policy evolution depends not only on isolated task wins, but on discovering task-appropriate mechanisms and refining policies under bounded feedback.

Physics-based Human-Scene Interaction (HSI) imitation learning is crucial for embodied intelligence as it bridges the gap between kinematic 3D motions and real-world dynamics. However, most existing methods focus on simplified scene settings, leaving complex environments largely unexplored, which limits their applicability in real-world scenarios. In this paper, we focus on HSI mimicry in complex environments. Under this complex setting, we observe an inherent trade-off between successfully performing interaction and maintaining natural, physically plausible motions. To address this challenge, we propose ComplexMimic, a framework that reconstructs diverse HSI by interpreting imperfect MoCap data. First, we introduce a Dual Flow Strategy, which learns two complementary experts: an imitation expert for accurate motion tracking and an interaction expert for collision-aware adaptation in complex scenes. Second, naive multi-expert distillation, which treats all experts equally, often under-samples challenging behaviors, limiting effective learning. To mitigate this issue, we propose a difficulty-aware distillation strategy that adaptively weights supervision and prioritizes hard-yet-learnable trajectories guided by failure statistics and learning progress signals. Extensive experiments on three benchmark datasets demonstrate that our approach outperforms current state-of-the-art methods.

As AI agents become increasingly capable of complex, long-horizon reasoning, rigorous and holistic evaluation is essential for measuring progress toward real-world healthcare applications. We introduce HealthAgentBench, a suite of 54 agentic healthcare tasks across 7 categories each with its unique environment. The benchmark suite spans diverse workflows throughout the patient journey and a broad range of modalities. Each task is designed to replicate an end-to-end clinical workflow: given minimal instructions, an agent must explore raw healthcare data, operate within a complex environment, and execute multi-step solutions that go beyond naive prompting. A final task success rate is reported to provide a single, interpretable metric for HealthAgentBench overall performance for each agent. Evaluating frontier agents on HealthAgentBench, we find that overall task success rate remains low, underscoring the difficulty of the suite. The strongest and the most cost effective agent, Codex GPT-5.5, achieves only approximately 42% success rate. Beyond aggregate performance, HealthAgentBench reveals nuanced strengths and weaknesses across task categories. Frontier agents show promise in automatically developing research modeling pipelines over EHR data, but medical imaging remains especially challenging, particularly for Claude Code models, while Codex GPT-5.5 shows emerging capability. Tasks that combine large search spaces with compositional reasoning requirements remain difficult for all current agents. Together, these results suggest that HealthAgentBench provides a challenging and realistic benchmark with substantial room for future progress. We release our benchmark at https://github.com/microsoft/HealthAgentBench.

WiFi sensing based on Channel State Information (CSI) promises ubiquitous, device-free perception, yet current research remains trapped in a Tower of Babel - fragmented into isolated silos where models are tailored to specific hardware dialects, fixed environments, and narrow tasks. The primary bottleneck is the Heterogeneity Gap: the disparity in signal dimensions, sampling rates, and semantic labels that prevents cross-system understanding. To bridge this gap, we propose a foundation-model framework that treats CSI not merely as raw signals but as a structured language with a learnable universal grammar. We first curate and standardize a large collection of heterogeneous real-world CSI datasets, establishing a unified infrastructure that allows incompatible signal formats to be treated as a single corpus. Second, we introduce a modular architecture that acts as a universal translator where lightweight dataset-specific adapters tokenize diverse signal inputs into a shared latent vocabulary, while a shared self-supervised Transformer backbone learns the temporal syntax of human motion and environmental dynamics. This design decouples sensing semantics from hardware syntax. Extensive evaluations show that by mastering this universal language, our approach consistently outperforms task-specific baselines and exhibits strong generalization capability in new environments, achieving superior efficiency in few-shot scenarios. By effectively absorbing heterogeneity, the framework offers a path toward robust, general-purpose wireless sensing, mirroring the linguistic generalization observed in Large Language Models. The code implementation is available at: https://github.com/cjychenjiayi/WiLLM.

Beryllium contamination surveys in radioactive areas are challenging for robots in environments cluttered with cables and electronics. To address this problem, we have developed a novel quadruped system augmentation: A lightweight, soft, and compliant tendon-actuated robotic tail mounted on a quadruped robot. The tail features a hollow, flexible backbone and a tendon-actuated soft gripper that enables the robot to pick up sampling tissues, swab contaminated surfaces, and release the tissues at designated collection locations for subsequent beryllium analysis. To enable intuitive teleoperation, a closed-form kinematic model and a singularity-robust task-space controller are developed. Experimental results demonstrate that gripper actuation has a negligible effect on robot shape, while common-mode tendon actuation provides an effective mechanism for stiffness modulation and preload control. Furthermore, experimental validation indicates that the proposed kinematic model provides a suitable basis for real-time task-space control. The proposed system combines the agility of legged locomotion with the compliance of soft robotic manipulation, enabling the complete contamination-survey procedure to be performed without human exposure. While motivated by beryllium contamination surveys at CERN, the proposed quadruped soft-tail concept is broadly applicable to legged robots operating in cluttered, confined, or hazardous environments where conventional rigid-link manipulators are undesirable.

Robust robot autonomy depends on scene representations that remain stable enough to support localization, navigation, and downstream decision making in dynamic environments. Monocular Gaussian Splatting SLAM provides high-fidelity mapping, but current uncertainty-aware methods still treat dynamic regions largely as per-frame observations. This makes the representation effectively memoryless: when a pedestrian slows, pauses, or reappears after occlusion, the current frame may look static, allowing dynamic content to be absorbed into the map and leaving persistent ghosting artifacts. We argue that this failure reflects a representation-level mismatch. Dynamic-ness is not an instantaneous appearance property, but a temporal property defined by motion history. Building on this view, we introduce Motion Permanence: the principle that an object's dynamic identity should persist over time rather than be re-decided from each frame independently. We realize this principle in MoPe, a memory-aware uncertainty filter for monocular Gaussian mapping. MoPe propagates the historical dynamic posterior through geometry-consistent SE(3) warping and fuses it with current-frame evidence using bounded Bayesian log-odds updates. The resulting persistent posterior guides tracking, mapping, dynamic-aware Gaussian insertion, and Gaussian-level post-cleanup. On Wild-SLAM, Bonn, and TUM sequences, MoPe improves tracking robustness and reduces residual ghosting, with the strongest gains on dynamic-human scenes that most directly violate the memoryless assumption. These results show that maintaining temporal dynamic state inside the scene representation is a practical step toward more reliable representation-centric autonomy in changing real-world environments.

Accurate localization of unmanned aerial vehicles (UAVs) is essential for applications such as structural health monitoring, especially in environments where Global Positioning System (GPS) signals are denied or unreliable, like indoor spaces, tunnels, urban canyons, or areas beneath large structures. To address this challenge, we propose Cross-Fusion, a novel method for real-time UAV localization that integrates data from a 3D Light Detection and Ranging (LiDAR) and a monocular camera. A key contribution is its cross-session fusion strategy, which integrates visual and geometric information collected from multiple agents during routine baseline surveys to improve localization consistency and map completeness. The system employs LiDAR-based odometry for motion tracking and image-based feature matching via a single red-green-blue (RGB) camera to correct drift and improve accuracy. Unlike visual-inertial systems, Cross-Fusion maintains a simple sensor setup and avoids the complexity of stereo or global shutter configurations. Experimental results demonstrate that Cross-Fusion achieves localization accuracy comparable to GPS-based methods and performs reliably in challenging feature-sparse environments.

Video generation models aspire to simulate dynamic environments, and several benchmarks now evaluate memory consistency across frames. However, most assess consistency only while the target remains in view, and the few that force objects out of view evaluate static scenes where nothing changes during occlusion. To bridge this gap, we introduce MemoBench, a diagnostic benchmark built around the disappear-and-reappear paradigm in dynamically changing environments: a target object undergoes a physical process, disappears from view, and must be correctly recovered in its updated state upon reappearance. We curate 360 ground-truth clips spanning synthetic and real-world scenes, and design an evaluation suite combining automated metrics with VQA-based assessment across four diagnostic pillars. Evaluation of eight state-of-the-art models reveals key insights and open challenges regarding memory consistency under the disappear-and-reappear paradigm.

Motion feasibility prediction plays a central role in robotics, particularly in task and motion planning and manipulation. A major bottleneck for this problem in cluttered environments is that infeasible planning attempts by Sampling-based motion planners (SBMPs) can incur substantial computational cost. Also existing approaches for infeasibility certification are limited to low-dimensional configuration spaces and often assume simplified geometric environments represented by primitive objects with known parameters. We study the complementary problem of learning motion feasibility prediction directly from raw RGB-D observations for a 7-DOF manipulator operating in realistic cluttered scenes. We introduce the first large-scale benchmark for this setting, comprising 2.7M grasp feasibility labels over 88 scanned objects and 190 cluttered tabletop scenes. We benchmark three representative classifier families spanning MLP- based, volumetric-CNN, and point-cloud-based Transformer architectures under matched training conditions. Our best model, GRASPFC-PTX (a point-cloud transformer), achieves an AUROC of 0.996 on Novel objects while providing predictions significantly faster than SBMPs.

Autonomous exploration in unknown environments requires unmanned aerial vehicles (UAVs) to efficiently generate informative trajectories while simultaneously constructing accurate maps. Although many existing exploration methods rely on geometric information, they often lack semantic awareness, resulting in suboptimal exploration efficiency and limited environmental understanding. To address this limitation, this paper proposes a semantic-aware exploration framework that adds semantic information to a roadmap-based exploration strategy. The proposed method builds on the Dynamic Exploration Planner (DEP), which incrementally constructs a Probabilistic Roadmap (PRM), and augments this roadmap with a semantic layer. A semantic reward function is introduced to prioritize regions containing meaningful objects and structures, enabling the UAV to focus on areas with higher information value. Furthermore, the roadmap is continuously updated to support efficient frontier selection and path planning during exploration. The proposed framework is implemented in ROS Noetic and Gazebo using an RGB-D sensor for simultaneous acquisition of geometric and semantic information. Experimental results in multiple simulated environments demonstrate that the proposed approach achieves exploration coverage rates between 90% and 94% while reducing exploration time and travel distance compared with conventional geometry-based exploration methods.

Facial landmark localisation is a prerequisite for developing automated, non-contact neonatal pain assessment methods. Clinicians use pain scales to judge the severity of pain, many of which rely on facial expression. However, facial landmark detectors trained on adult faces perform poorly in neonatal clinical environments due to frequent occlusions caused by medical equipment, varied head poses, and challenging imaging conditions, including motion blur triggered by sudden pain-related movements. We propose an end-to-end facial landmark detector capable of predicting 68 landmarks on neonatal faces in clinical environments. We combined 37,459 single-face images from 11 public datasets, standardised to 68-point markup, with 1,123 manually annotated frames from a neonatal research dataset (totalling over 76,000 landmarks). A YOLO-based keypoint model was adapted to regress the facial landmarks, initialised with weights from a pretrained neonatal face detector. On public datasets, our proposed model achieved state-of-the-art performance: Normalised Mean Error (NME) = 5.37, Failure Rate (FR) = 12.5%, Area Under the Cumulative Error Curve (AUC) at AUC0.08 = 38.00% and AUC0.1 = 48.70%. On the clinical neonatal test set, before fine-tuning, the model achieved the lowest Detection Failure Rate (DFR) = 5.3% among all baselines and showed strong generalisation. After fine-tuning, performance improved further to NME = 6.36, FR = 22.30%, DFR = 1.77%, AUC0.08 = 29.24% and AUC0.1 = 40.25%. To the best of our knowledge, this represents the first end-to-end 68-point neonatal facial landmark detection model. With further dataset expansion and refinement, it could support downstream tasks in neonatal health monitoring and pain-related facial analysis.

Reliable localisation of the neonatal face is the first step for several video-camera based non-contact assessments such as pain and distress related facial expression analysis, pain scoring, cardiorespiratory signal extraction and cessation of breathing alerts. However, major challenges persist in neonatal clinical environments. Cluttered backgrounds, illumination changes and poor lighting conditions can reduce the accuracy of face detection models. Clinical interventions, monitoring equipment and, in some cases, medical devices can obstruct the face, making visual assessment difficult. We propose a one-stage YOLOv11m-based model tailored for face detection of infants in neonatal clinical environments. We combined multiple publicly available datasets (VGGFace2, CelebA, FDDB, WIDER FACE) to train and evaluate our proposed model. We then fine-tuned our model on a neonatal research dataset involving 228 videos from 114 recording sessions of 113 independent infants. Before fine-tuning, our model achieved an AP50 of 0.87, surpassing the performance of three state-of-the-art general face detectors. Performance improved further to an AP50 of 0.96 after clinical-domain adaptation. Evaluating face detection performance across different datasets remains a challenge due to the lack of publicly available neonatal datasets. Prioritising the creation of such datasets, while upholding appropriate privacy safeguards and ethical standards in their creation and use, would greatly support further progress in this field.

Visual navigation in unknown environments remains a core challenge in mobile robotics, especially for resource-constrained platforms. Most existing approaches rely on loosely coupled modular pipelines and strong assumptions on perception quality or environmental structure, often resorting to multi-modal sensor suites that increase system complexity and deployment cost. Vision-only navigation offers a lightweight alternative, but its performance degrades severely under motion blur, low texture, and illumination changes, largely because they neglect the tight coupling between commanded motion and perception. While perception-aware methods partially address this issue, they typically optimize individual modules and fail to propagate uncertainty consistently across the navigation stack. In this paper, we present UNSEEN, a unified uncertainty- and perception-aware navigation framework that explicitly couples localization, mapping, and planning using only a front-mounted camera. UNSEEN estimates sparse maps and robot poses with associated uncertainties at 6Hz, and leverages them to plan trajectories that jointly optimize task progress and estimation accuracy in receding-horizon. Simulations and extensive real-world experiments in unknown environments demonstrate the robustness of the proposed approach, with UNSEEN-SLAM reducing absolute translational error by 9.8% and UNSEEN-Plan improving estimation accuracy by up to 45% compared to state-of-the-art methods, while achieving a 100% task success rate.

Deep research agents have demonstrated remarkable capabilities in autonomous information gathering and synthesis, yet their training remains constrained by the static nature of simulated environments, the limits of fact-retrieval-only task designs, and the inefficiency of outcome-based reinforcement learning. In this work, we propose MetaResearcher, a novel framework that scales deep research agent training across four synergistic dimensions. First, we introduce an Evolving Virtual World that injects temporal dynamics and adversarial misinformation into the training environment, forcing agents to develop source credibility assessment and temporal conflict resolution skills. Second, we design Discovery-Oriented Tasks -- including hypothesis generation and contradiction resolution -- that transcend simple fact retrieval and push agents toward genuine research behaviors. Third, we propose a Self-Reflective Meta-Reward mechanism within the GRPO framework that jointly optimizes for answer correctness, search path efficiency, reflection depth, and tool call diversity, directly addressing the repetitive action loop problem observed in prior work. Fourth, we introduce a Heterogeneous Multi-Agent Swarm architecture comprising specialized Scout, Filter, and Synthesizer models that learn collaborative research strategies through coordinated reinforcement learning. Built upon the LiteResearcher infrastructure, MetaResearcher requires zero marginal API cost for training while targeting substantial improvements in both benchmark performance (GAIA, Xbench-DS) and epistemic robustness under adversarial conditions. We present the complete framework design, training methodology, and planned experimental validation.

Training embodied agents in the real world requires skilled operators and expensive hardware. Simulation environments offer a compelling alternative by enabling large-scale, cost-effective data augmentation. Consequently, rapidly constructing high-fidelity simulation scenes with a minimal sim-to-real gap has become a critical objective in robot learning. While reconstruction-based methods provide superior visual quality, current workflows are hindered by inefficient data acquisition and subpar foreground object extraction. We thus propose GASE, a highly automated system for simulation scene construction. GASE leverages multi-view video streams from panoramic camera arrays to enable rapid environment scanning. To ensure high-quality asset generation, our pipeline introduces a camera-pose-based strategy that robustly extracts objects across frames in the 2D domain, followed by high-fidelity scene inpainting. Foreground objects and the static background are then reconstructed independently and seamlessly imported into physics simulators for policy training. Extensive experiments demonstrate that GASE outperforms existing 3D Gaussian-based methods in segmentation accuracy by over 10% while achieving state-of-the-art inpainting quality. Furthermore, real-robot deployments across manipulation and navigation tasks maintains a performance gap of less than 10% compared to policies trained purely on real-world data. These results confirm that GASE provides an efficient and highly effective solution for bridging the sim-to-real gap. Code will be released.

Large language model (LLM) agents have made rapid progress on short-horizon, well-scoped tasks, yet their ability to sustain coherent decisions in dynamic long-horizon environments remains uncertain. We introduce RetailBench, a data-grounded simulation benchmark for evaluating tool-using LLM agents in single-store supermarket operation. RetailBench models retail management as a partially observable decision process and is designed to support thousand-day-scale simulations. In this environment, agents must manage pricing, replenishment, supplier selection, shelf assortment, inventory aging, customer feedback, external events, and cash-flow constraints. We evaluate seven contemporary LLMs under representative agent frameworks over a 180-day evaluation horizon and compare them with a privileged oracle policy. Results show substantial variation across models: only a small subset survives the full evaluation horizon, and even the strongest LLM runs remain substantially behind the oracle policy in final net worth and sales outcomes. Behavioral analysis attributes these gaps to incomplete evidence acquisition, surface-level decision making, and the lack of a consistent long-horizon policy. RetailBench provides a controlled testbed for studying reliable autonomy in economically grounded long-horizon decision-making.

We present AliyunConsoleAgent, a web agent framework for automated documentation verification in real-world cloud consoles. Major cloud platforms encompass hundreds of products with rapid feature iteration, causing console UIs to frequently diverge from their corresponding documentation. Verifying that documented procedures accurately reflect the current console and can be executed end-to-end demands an estimated 4 million recurring inspections annually, yet manual coverage remains below 1%. While agent systems built on frontier proprietary models achieve high success rates, their prohibitive cost and data privacy constraints preclude large-scale deployment. We propose a two-stage training paradigm: supervised fine-tuning (SFT) on distilled frontier-model trajectories, followed by reinforcement learning using Group Relative Policy Optimization (GRPO) and a dual-channel outcome reward model in real cloud environments. To support large-scale RL training, we construct a high-determinism rollout system featuring Terraform-based resource pre-provisioning and LLM-driven on-demand provisioning, which effectively isolates environment noise from the training signal. We further introduce a rule-based reward evaluation protocol grounded in backend audit logs, providing objective, reward-hacking-resistant outcome judgment. Our model evolves from mechanical instruction following to autonomous decision-making with cloud console and product-specific understanding. Experiments on a challenging 278-task benchmark where the best frontier model achieves only 65.34% demonstrate that AliyunConsoleAgent-32B achieves a 63.52% mean success rate -- a 20.24 percentage-point improvement over the base model, narrowing the gap to the best frontier proprietary model to 1.82 pp (bootstrap 95% CI [-1.27, 7.39]) -- at 92% lower inference cost.

The dynamic nature of Internet of Things (IoT) environments affects the long-term effectiveness of Machine Learning as a Service (MLaaS) compositions. Existing adaptive composition methods are mainly based on service replacement or re-composition, where identifying suitable substitutes is difficult and time-consuming. To address this, we propose a novel Test-Time Adaptive (TTA) composition framework for MLaaS in IoT environments. First, we introduce a TTA-aware composability model to determine whether adapted services remain compatible with the existing composition. Next, we design a service-level adaptation model to adjust individual services during inference while preserving composition performance. Experimental results demonstrate that the proposed framework reduces computational time more effectively than traditional adaptive approaches.

Training LLMs to orchestrate multi-step tool calls is held back by three coupled obstacles: realistic stateful execution environments are costly to build, synthetic training queries are often detached from the server's actual state (so the generated tool calls fail to execute), and recall-based RL rewards incentivize verbose tool-calling patterns. We present PROVE (Programmatic Rewards On Verified Environments), a framework with three contributions: (1) a library of 20 stateful MCP (Model Context Protocol) servers exposing 343 tools, enabling live-execution RL training with session-scoped state isolation; (2) a state-machine data synthesis pipeline that generates multi-turn tool-call trajectories grounded in live-sampled server state, so generated queries reference entities that actually exist; and (3) a multi-component programmatic reward with an adaptive efficiency penalty that counters the verbosity incentive of recall-based rewards. We train four models (Qwen3-4B, Qwen3-8B, Qwen2.5-7B, Granite-4.1-8B) with GRPO on the resulting ~13K training examples. On BFCL Multi-Turn, tau2-bench, and T-Eval, PROVE yields improvements of up to +10.2, +6.8, and +6.5 points respectively, demonstrating that this framework yields consistent gains on multi-step tool orchestration across two model families.

Structured financial audit verification is difficult for language-model agents because correctness depends on structured evidence rather than text alone. A model must link reported facts to taxonomy concepts, traverse calculation or dimensional relations, and recompute expected values before applying an audit rule. We propose AuditFlow, a graph-grounded multi-agent framework that separates adaptive search from deterministic verification. AuditFlow builds a symbolic environment from a static US-GAAP taxonomy graph and a dynamic XBRL filing graph, and exposes it through typed tools for fact retrieval, taxonomy traversal, numerical checking, and rule evaluation. Two junior auditors inspect each case from regulatory and evidentiary views, while a senior auditor resolves disagreements and can request further investigation. The final reports are fused through evidential aggregation to produce an audit verdict, expected value, evidence trail, and trustworthiness score. On a FinAuditing-derived FinMR sample, AuditFlow reaches 82.09% joint audit accuracy under GPT-5.5, outperforming the strongest baseline by 14.93 points. Removing deterministic checks drops accuracy to 17.91%, showing that the symbolic environment performs the verification step that the model cannot reliably replace.