Search arXiv⌕ Search

arXiv subjects

Li Yang

Publications and source records attributed to Li Yang.

At least 19 recordsLinked to original sources
arXiv2026

SD3: The First Operationally-Proven, Fully Heterogeneous-Accelerated Global Kilometer-Scale Model with Inline Physics Architecture

We present Super Dynamics on the Cube (SD3), the first operationally-proven, fully heterogeneous-accelerated global kilometer-scale model. Through a full-stack reconstruction in native CUDA C, SD3 introduces an "Inline Physics" architecture that deeply couples its non-hydrostatic cubed-sphere dynamical core with fast physical processes. This design maintains high physical consistency, ensures energy and mass conservation, and mitigates "Grey Zone" mismatches. Five years of continuous operational forecasting, together with validation against diverse observations, demonstrates the scientific fidelity of SD3. Recently, SD3 achieved an unprecedented 205.7 simulated days per day at global 1.5 km resolution using 24576 GPUs. SD3 establishes a hardware-agnostic, portable, and reproducible paradigm across diverse Exascale ecosystems. Beyond weather and climate research, SD3 offers a transformative paradigm for the HPC community and the next-generation Digital Twin Earth to establish a robust, high-performance blueprint for the future of Earth system modeling.

physics.ao-ph↗
arXiv2026

UniManip: Open-World Robotic Manipulation via Agentic Operational Graphs

Achieving general-purpose robotic manipulation requires robots to bridge high-level semantic intent with low-level physical interaction in unstructured environments. However, zero-shot generalization remains challenging: end-to-end Vision-Language-Action (VLA) models often lack the precision required for long-horizon tasks, while traditional hierarchical planners suffer from semantic rigidity under open-world variations. We present UniManip, a unified framework centered on an Agentic Operational Graph (AOG), an explicit cognitive and operational backbone for task orchestration, state tracking, execution verification, and closed-loop failure recovery. High-level semantic intents are grounded into actionable primitives via a physically grounded scene graph that provides geometric affordances, contact-aware interaction parameters, and collision-free trajectory synthesis. By coupling a high-level Agentic Layer for task orchestration with a low-level Scene Layer for dynamic state representation, UniManip continuously aligns abstract planning with geometric constraints, enabling robust execution. Unlike static pipelines, UniManip operates as a dynamic agentic loop: it instantiates object-centric scene graphs from unstructured perception, parameterizes these representations into collision-free trajectories via a safety-aware local planner, and exploits structured memory to autonomously diagnose and recover from execution failures. UniManip achieves state-of-the-art performance on LIBERO-Pro, while extensive real-world experiments demonstrate robust zero-shot generalization across diverse tasks, embodiments, and dexterous hands, outperforming leading VLA models and hierarchical baselines by 22.5% and 25.0% in task success rate, respectively. Open-source project page: https://henryhcliu.github.io/unimanip.

cs.RO↗
arXiv2026

Central limit theorem for the Smoluchowski--Kramers approximation

We establish the central limit theorem (CLT) for Smoluchowski--Kramers approximation of Langevin equations with state-dependent, possibly non-symmetric friction and multiplicative noise. To the best of our knowledge, this is the first such result, even in the case of constant friction and additive noise. The result differs in an essential way from classical averaging CLTs and reveals a new two-scale fluctuation phenomenon. The normalized difference process retains a persistent fast component and is generally not tight in $C([0,T];\mathbb{R}^d)$, thus the usual continuous-path CLT cannot hold. We derive an exact decomposition of the normalized difference into an accumulated slow fluctuation and a rapidly relaxing velocity correction. The auxiliary slow fluctuation converges to a linear diffusion driven by both the original Brownian motion and an additional independent Brownian noise whose effective covariance is characterized explicitly through a quadratic Poisson corrector. For every fixed $t>0$, the fast correction converges conditionally to its local Gaussian equilibrium. Consequently, the fixed-time fluctuation limit is identified as a Gaussian mixture: $$ \frac{\bar X_t-X_t^\varepsilon}{\sqrt{\varepsilon}} \Longrightarrow \bar Y_t+N^{1/2}(\bar X_t)ξ, $$ where $\bar X_t$ is the overdamped limit, $\bar Y_t$ is the limiting slow fluctuation, $N$ is the covariance of the terminal velocity correction and solves a Lyapunov equation, and $ξ$ is a standard Gaussian random variable independent of $(\bar X_t,\bar Y_t)$. A quantitative convergence rate is also provided. The result fills the remaining gap in the asymptotic theory of the overdamped approximation at the central limit scale, complementing the existing law of large numbers, moderate deviation and large deviation results.

math.PR↗
arXiv2026

FlowVLA: Learning Flow-Aware Physical Completion for Quantitative Robotic Pouring

During steady pouring, images and robot pose may remain nearly unchanged while transferred mass accumulates, making the return decision ambiguous for a whole-task vision-language-action (VLA) policy. We present FlowVLA, which separates motor generation from physical-completion estimation. A shared $π_{0.5}$ action expert learns approach, pour, and return with boundary-aware supervision. A target-independent estimator fuses visual, joint, wrist-wrench, and recurrent force-belief histories to maintain an estimate of transferred mass. The controller initiates return when this estimate plus predicted residual outflow reaches the requested mass. In coffee-pot/water experiments, FlowVLA achieves 13.1 g goal-balanced mean absolute error (MacroMAE), lower than all tested VLA baselines. Across all attempts, 53.7% both complete the task and finish within +/-10% of the requested mass. The same model interpolates to held-out 150/250-g targets without retraining. Independently trained models further evaluate the formulation in three additional container-fluid settings.

cs.RO↗
arXiv2026

Recording-Aware Evaluation of Multimodal OT Intrusion Detection

Machine-learning-based intrusion detection systems (IDSs) for operational technology (OT) are commonly evaluated on benchmark datasets in which normal and attack data are collected in separate recordings. In such settings, high performance may reflect recording-associated differences as well as attack behavior. We introduce a recording-aware evaluation procedure that compares attack windows with background from the same recording, uses prior data for model fitting and calibration, tests temporal alignment against randomized event schedules, and reports event coverage together with alarm burden. On the ProvICS multimodal dataset, cyber modalities strongly distinguish the benign and attack recordings before the first labeled attack. Same-recording evaluation reverses the apparent modality ranking across all studied window sizes: the physical view becomes the strongest individual modality, while cyber-only fusion shows the largest difference between cross-recording and same-recording AUC, reaching 0.54. Operational analysis further shows that permissive event-overlap metrics can overstate detection when a small number of long alarm episodes cover multiple actions. An external analysis on HAI shows smaller and reference-dependent recording sensitivity. These findings show that cross-recording benchmark performance should be interpreted alongside same-recording, forward, temporal, and operational evaluations.

cs.CR↗
arXiv2026

Online AutoML: Evaluating Poisoning Attacks on Cost-Utility Adversarial Training Defence Strategy for IoT Networks

Machine learning (ML)-powered poisoning attack vectors are adversarial maneuvers whereby an attacker intentionally inserts, corrupts, or alters training data to distort an ML model's learning process. The objective is to diminish model efficacy, instill biases, induce misclassifications, or include concealed backdoors that may be attacked during implementation. In streaming contexts, poisoning attacks pose significant risks since models perpetually update based on incoming streams of data. An assailant may incrementally introduce harmful samples into this data stream, leading the model to assimilate erroneous features over time without timely identification. Therefore, this study is aimed at evaluating the efficacy of the adversarial training (AT) defense approach against poisoning attacks (label flip and noise injection) using an online AutoML pipeline for Internet of Things (IoT) networks. Specifically, poisoning attacks (label flip and noise injection) were applied to streaming-capable AutoML learners (Hoeffding Tree (HT), Leveraging Bagging (LB), Adaptive Random Forest (ARF), Hoeffding Adaptive Tree (HAT), and Streaming Random Patches (SRP)). Under the strongest poisoning rate (PR = 1.0), AT-SRP achieved the highest F1-score against label flip poisoning (0.904), while AT-LB achieved the highest F1-score against noise-injection poisoning (0.933). Finally, several drift detection methods were used for rolling accuracy and prequential evaluation.

cs.CR↗
arXiv2026

Instance-anchored interaction evidence: Grounding robot plans in human pointing and handling

A robot that assists people must often act on what a person has shown rather than said: which of several identical cartons was pointed at, or which box was handled. The plan is executed from the final scene, whereas the evidence occurs earlier, possibly on objects that have since moved. We propose instance-anchored interaction evidence (IAE), which registers every object of the final scene to its public identifier, keeps each identity through the video by backward mask propagation, and describes every frame by the geometry between hands, forearms and these instances. An evidence network trained only from task outcomes scores the instances. For pointing tasks, a grammar-constrained dynamic program trained with a structured loss decodes object-destination programs; symbolic programs handle reference disambiguation and, without learning, episodic tasks. On 1,255 WatchAct benchmark requests, scored by symbolic execution, IAE reaches 64.2% plan success on implicit-intent tasks against 27.5% for a 32B vision-language model (strict success 49.7% against 15.4%), and 46.4% against 27.0% on restoration, reversal and imitation without task-specific training. Controls with the same perception overlays, the same 32 frames, forward tracking, or a relation model trained on the same labels do not explain the gain. Given IAE's evidence as text with its meaning explained, the same language model reaches 57.4%: most of the gain comes from the instance-anchored evidence, and the explicit programs add 6.8 points at a fraction of the cost. Pointing remains the hardest case, with 16.9% strict success. The code is available at https://github.com/WeiZhou96/iae-watchact.

cs.RO↗
arXiv2026

Fixed-Reference Pose Residuals for Measuring Cross-Dataset Cue Transfer in Human-Robot Interaction Anticipation

Social and service robots in public spaces need to anticipate which nearby person is about to approach and touch them, so that a response can be prepared before contact. It is largely unknown which cues support this anticipation when a model trained with one robot is used on another robot at a different site. We study this question with a fixed-reference pose residual (FRPR) model: a geometry predictor built from the person's bounding box and mask is trained and frozen, and a temporal network then learns from body pose an additive correction to its logit, so that every prediction splits exactly into a geometry term and a pose term. Between two public egocentric datasets recorded by different robots, HUI360 and SSUP-A, with every choice made on source data, the pose correction raised average precision (AP) from 0.277 to 0.321 from SSUP-A to HUI360 and gave no measurable gain in the opposite direction; the same asymmetry held over a stronger, source-selected geometry reference. Freezing gave no AP advantage over joint training, and simple geometric baselines and tree ensembles remained competitive or better, so the construction serves measurement rather than prediction accuracy. A head-orientation residual added small gains in both directions. Post hoc, whether a person faces the camera kept its discriminative direction across datasets, whereas head pitch reversed. With thresholds chosen on source data, the neural models that use geometry detected at most 17% of target interactions. Code and processed data are available at https://github.com/WeiZhou96/FRPR-interaction-anticipation.

cs.RO↗
arXiv2026

Evasion Attacks on Cost-Utility-Based Adversarial Training for Online AutoML in IoT Networks

As Internet of Things (IoT) networks increasingly depend on machine learning for anomaly, malware, intrusion detection, and network monitoring, such systems have become attractive targets for evasion attacks. Evasion attacks pose a major security risk because an adversary intentionally modifies input data to mislead a trained model into producing incorrect predictions while evading detection. This study evaluates the impact of black-box evasion attacks on a cost-utility-based adversarial training defense strategy in an Online AutoML context for IoT networks. Specifically, evasion attacks were applied to online learners, including Hoeffding Tree (HT), Leveraging Bagging (LB), Streaming Random Patches (SRP), Hoeffding Adaptive Tree (HAT), and Adaptive Random Forest (ARF). By developing naive and adversarially trained (AT) versions of these online learners, we generated clean and adversarial accuracies for each model. The results show that the AT versions of LB and SRP performed best, achieving the highest adversarial accuracy (0.985) and high clean accuracy (0.993) at the highest cost budget of 1.00, with a maximum accuracy reduction of only 0.8%. Finally, drift detection was conducted using the Early Drift Detection Method (EDDM).

cs.CR↗
arXiv2026

Electrical Probing of Dark Excitons through Microwave Permittivity

Excitons in atomically thin semiconductors are almost always probed through their optical signatures, because the short lifetimes of these transient quasiparticles are generally assumed to preclude electrical detection. Here we show that photoexcited excitons in monolayer tungsten disulfide produce a large, optically tunable permittivity at gigahertz frequencies, and that the effect provides a contact-free electrical route to imaging dark excitons at the nanoscale. Using laser-illuminated microwave impedance microscopy, we find that high-purity encapsulated flakes exhibit a purely dielectric response resonant with the exciton spectrum, whereas defect-rich samples are governed by conventional photoconductivity. Spatial mapping of diffusion and sublinear power dependence identify long-lived dark excitons as the dominant contributors, and first-principles modelling of exciton polarizability reproduces the measured susceptibility. Our results establish excitons as optically tunable dielectric elements and introduce microwave microscopy as a direct electrical probe of dark-exciton transport with sub-100 nm resolution.

cond-mat.mes-hall↗
arXiv2026

An LLM-Assisted AutoML Framework for Intrusion Detection in IoT Networks

Internet of Things (IoT) systems are increasingly deployed in smart homes, transportation, energy systems, and critical infrastructure. This broad connectivity improves service intelligence, but also enlarges the attack surface of IoT networks. Machine Learning (ML)-based Intrusion Detection Systems (IDSs) are widely used to identify malicious network threats and protect IoT systems, but developing effective ML-based IDS models often requires human expertise and repeated manual decisions on many procedures, including data pre-processing, feature selection, model selection, and hyperparameter tuning. Automated Machine Learning (AutoML) reduces this burden by automating steps of the ML pipeline using optimization techniques, but conventional AutoML methods can consume substantial optimization time because they explore broad candidate model families and large hyperparameter spaces. This paper proposes a Large Language Model (LLM)-assisted AutoML framework for IoT intrusion detection. The proposed framework uses an LLM as a policy generator that converts dataset profiles into bounded and validated AutoML policies for automated data balancing, automated feature engineering, and Combined Algorithm Selection and Hyperparameter Optimization (CASH). Under an equal 10-trial budget, the proposed LLM-assisted policy achieves higher weighted test F1-score than traditional AutoML using the Tree-structured Parzen Estimator (TPE) on both datasets, reaching 99.680% on CICIDS2017 and 99.186% on IoTID20. Relative to the broader 30-trial Traditional AutoML-TPE baseline, the 10-trial proposed method reduces optimizer time by 63.7% and 49.9%, respectively, while achieving slightly higher F1-score. These results show that a bounded LLM policy can improve the quality of a low-budget AutoML search while retaining a clear efficiency advantage relative to a larger conventional search budget.

cs.CR↗
arXiv2026

Multimodal Aspect-Level Sentiment Analysis Based on Gated Noise Filtering and Emotion-Relevance Interaction

Multimodal Aspect-Based Sentiment Analysis (MABSA) infers fine-grained sentiment polarity toward specific aspects by jointly modeling text and images. Despite progress in cross-modal fusion, two challenges remain in multi-aspect settings: (1) multimodal noise, where aspect-irrelevant content distracts sentiment learning; and (2) weak cross-modal sentiment alignment, as visual evidence can be ambiguous and textual--visual sentiments may conflict, limiting multimodal complementarity. To address these issues, we propose a Gated Noise-filtered Sentiment-Relevance Interaction (GNSRI) framework. It employs a gated noise-filtering module to suppress sentiment-irrelevant features and enhance aspect-aware sentiment cues, and a sentiment-relevance interaction module to capture consistent and conflicting cross-modal signals at micro and macro levels. Finally, a learnable decision fusion mechanism adaptively combines predictions from textual, visual, and cross-modal branches at the aspect level. Experiments on public MABSA benchmarks show that GNSRI outperforms state-of-the-art methods, improving accuracy by 1.94\% and 2.06\% on Twitter-2015 and Twitter-2017, respectively.

cs.MM↗
arXiv2026

CacheTrap: Unveiling a Stealthier Gray-Box Trojan against LLMs

The rapid advancement of large language models (LLMs) has sparked growing interest in understanding their security vulnerabilities, particularly Trojan attacks that enable stealthy manipulation of model behavior. Traditional Trojan methods typically alter inputs and/or model weights, relying on white-box assumptions that require access to data or model internal parameters. In this work, we present CacheTrap, the first gray-box Trojan attack targeting the Key-Value (KV) cache of LLMs. This method induces a single-bit flip in the KV cache, serving as a transient trigger. When activated, this trigger causes the model to exhibit targeted actions without changing inputs or model weights. CacheTrap introduces an efficient search algorithm to locate vulnerable positions in the KV cache, independent of model weights or datasets. Extensive experiments on five open-source LLMs show a remarkable near 100% attack success rate (with the trigger) while preserving benign accuracy (without the trigger) by flipping just one bit in the KV cache. Code is available at https://github.com/ML-Security-Research-LAB/CacheTrap

cs.CR↗
arXiv2026

Moiré Topology in Twisted Structures with Noncollinear Spin-Orbit Coupling

Moiré superlattices provide a powerful platform for flat bands and correlated topological phases, yet most established examples rely on valley-contrasting Berry curvature in hexagonal lattices. Here, we propose a different route to achieve topological moiré minibands based on noncollinear spin orbit coupling in centrosymmetric type II SOC bilayers. Interlayer hybridization opens local pseudogaps and produces sharply localized Berry curvature, which twisting reconstructs into isolated topological minibands without requiring valley degrees of freedom or hexagonal symmetry. We demonstrate this mechanism in tetragonal Dresselhaus-SOC HgI2, where a Lieb-like moiré potential yields topological flat bands. To improve miniband isolation, we develop a physics-informed machine-learning surrogate that identifies stronger SOC as a key design principle and guides the replacement of Hg by Pb. The resulting PbI2 minibands are narrower and better isolated, supporting correlation-driven magnetism and tunable quantum spin Hall and Chern insulating phases, thereby providing an optimized material realization for experimental exploration.

cond-mat.mtrl-sci↗
arXiv2026

Infinite-horizon time-inhomogeneous Kolmogorov equation and non-autonomous multi-scale stochastic systems

We propose a unified approach, based on the infinite-horizon time-inhomogeneous Kolmogorov equation with a parameter, to study the asymptotic behavior of non-autonomous multi-scale stochastic systems with irregular coefficients, where both the fast and the slow equations depend on the highly oscillating time component. We establish both the weak and strong convergence of double averaging principle as well as the functional central limit theorem with new homogenized coefficients given in terms of the evolution system of invariant measures of the time-inhomogeneous fast frozen equation. Moreover, rates of convergence are obtained as byproducts of our argument, which do not depend on the regularities of the coefficients with respect to the time component and the fast variable.

math.PR↗
arXiv2026

Dynamic-Robust Photometric-Semantic Reconstruction for Open-Vocabulary 3D Scene Understanding

The integration of novel view synthesis (NVS) and open-vocabulary segmentation (OVS) has recently yielded powerful feed-forward 3D foundation models. However, their inherent reliance on static-scene assumptions leads to severe misalignment of spatial features in unconstrained dynamic environments. To bridge this critical gap, we propose SPAR, a novel joint semantic-geometric encoding architecture that explicitly isolates transient dynamic noise prior to latent space aggregation. Furthermore, we introduce a dynamic-region-aware end-to-end training paradigm that structurally couples motion estimation with multi-view visual and semantic learning. This unified approach enables the network to inherently resolve motion conflicts and distill multi-view consistent, temporally stable scene representations from dynamic inputs. Extensive experiments on the challenging D-RE10K benchmark demonstrate that SPAR achieves state-of-the-art performance. Our end-to-end approach achieves exceptional novel view synthesis quality, yielding a PSNR of 22.15 dB and 23.33 dB given only 3 and 4 input views respectively. Despite being trained in a self-supervised manner, our model achieves an mIoU of 88.5% for motion mask prediction. Furthermore, our analysis reveals a strong inter-task synergy between photometric scene reconstruction and semantic understanding, where semantic synthesis learning consistently enhances photometric fidelity in novel view rendering. Code will be available at https://github.com/dmucby/SPAR.

cs.CV↗
arXiv2026

Robustness of Anomaly Detection Models for Industrial Control Systems under Training-Time Data Contamination

Machine-learning-based anomaly detection is increasingly used in industrial control systems (ICS), yet most studies assume that detector training data is trustworthy. In practice, training data may be corrupted through compromised logs, labeling errors, manipulated historian records, or unsafe retraining processes. This paper evaluates the robustness of offline ICS anomaly-detection pipelines on the Secure Water Treatment (SWaT) benchmark under training-time contamination. We assess 11 heterogeneous anomaly detectors under three contamination strategies: random injection, similarity-targeted injection, and feature-noise injection. The first two insert attack samples into the nominal training pool, while the third adds bounded Gaussian noise to selected normal training samples. These attacks are contamination-based rather than gradient-driven poisoning methods. Contamination budgets from 1% to 10% are evaluated using clean validation and test sets under a unified offline protocol. The results show that robustness is strongly model-dependent and cannot be predicted from clean-data performance alone. Injection-based contamination causes the greatest degradation, particularly for local-density and distance-based detectors, whereas feature-noise contamination has a comparatively limited effect. PCA, SVM, HBOS, and IForest remain relatively stable, while the tuned neural detectors demonstrate intermediate robustness. Overall, the findings highlight the importance of training-data integrity in ML-enabled ICS monitoring, subject to the evaluated dataset, models, and threat assumptions.

cs.CR↗
arXiv2026

Anomalous stabilization of excitons by metallic proximity

Metallic environments are generally expected to suppress excitons through strong dielectric screening, yet their influence can differ in composite systems where metallic and insulating regions coexist. Here, we investigate excitonic states in PdxCu1-xCrO2 thin films across a percolation-driven metal-insulator transition. Optical spectroscopy and many-body GW calculations show that CuCrO2 hosts strongly bound excitons with a binding energy of about 489 meV. With increasing Pd substitution, the system approaches an insulator-to-metal transition near x = 0.5, consistent with the site-percolation threshold of a triangular lattice. In the pre-percolation regime, the excitonic resonance redshifts by 241 meV while the Tanguy continuum onset remains nearly unchanged, consistent with a substantial increase in exciton binding energy before metallization. An image-charge-based excitonic hydrogen model shows that isolated metallic regions can enhance electron-hole binding through image-charge interactions, whereas conventional screening is recovered once a continuous metallic network forms. Although this model provides a possible interpretation of the observed excitonic evolution, an alternative scenario in which metallic and excitonic responses originate from electronically distinct states and evolve independently cannot be excluded. These results reveal unusual metallic-excitonic coexistence near a percolation-driven metal-insulator transition and suggest nanoscale metallic proximity as a possible route for modifying excitonic interactions.

cond-mat.str-el↗