arXiv subjects

Nan Deng

Publications and source records attributed to Nan Deng.

At least 19 recordsLinked to original sources

arXiv2026

China's Shrinking Home Bias and Rising Disruptive Impact: Evidence from a Global Citation Network Analysis

China has become the world's largest producer of scientific publications, yet concerns persist that this growth is inflated by excessive domestic citation practices. In this study, we analyze a citation network of over 45 million publications from Web of Science (1980-2025) to investigate China's home citation bias and research impact. Using a network reshuffling null model to control for the structural effect of publication volume, we find that China's home citation bias is less pronounced than commonly assumed and has been steadily declining over the past two decades. Chinese researchers do not exhibit a significantly stronger home citation preference than other major countries, indicating increasing internationalization rather than insularity. Furthermore, using the persistent disruption framework, we show that Chinese papers are converging toward American papers in their capacity to produce paradigm-shifting work. These findings challenge prevailing narratives about Chinese scientific home bias and suggest that China's advances in research impact.

physics.soc-ph↗

arXiv2026

Uniform-in-time strong convergence rates of fully discrete approximations for stochastic Cahn--Hilliard equations with multiplicative noise

This paper investigates the uniform-in-time strong convergence rates of a fully discrete approximation for the stochastic Cahn--Hilliard equation driven by multiplicative noise in spatial dimensions $d\in\{1,2,3\}$. The proposed scheme combines a spectral Galerkin method in space with a backward Euler scheme in time. The main analytical difficulties arise from the state-dependent stochastic perturbation, the absence of a global monotonicity structure for the nonlinear term, and the fourth-order nature of the Cahn--Hilliard operator. In particular, these features make the derivation of uniform $L^{\infty}$-moment estimates highly nontrivial in three dimensions. For the continuous equation, by utilizing the It\^{o} formula to $\|u\|^p$ and introducing the energy functional $\mathcal{E}(u(t))$, we derive the uniform moment boundedness of the solution. At the fully discrete level, we develop discrete energy estimates and close the required high-order moment bounds through an induction argument. Based on these regularity estimates, we deduce uniform-in-time strong convergence rates for the fully discrete scheme. Moreover, we prove the existence and uniqueness of invariant measures for both the exact dynamics and the fully discrete numerical dynamics. Numerical experiments are provided to confirm the theoretical findings.

math.NA↗

arXiv2026

Your Agentic LLMs Secretly Encode Indirect Prompt-Injection Exposure in Hidden States

Agentic LLMs are vulnerable to indirect prompt injection (IPI) attacks, e.g., malicious side-tasks hidden in external tool results. While many efforts have sought to address this threat, little is known about the internals of agentic LLMs when they are exposed to IPI attacks. For simplicity, we refer to this condition as IPI exposure. In this paper, we study IPI exposure from three perspectives. (1) Probing: Across eight models, including the 753B-parameter GLM-5.2 and the 2.8T-parameter Kimi-K3, simple linear probes trained on pre-generation hidden states can predict LLMs' IPI exposure. These probes achieve 0.90+ AUROC on unseen attacks, agent instructions, and task suites; they remain robustly predictive under adaptive attacks and in cross-lingual settings. (2) Defense: We reveal and diagnose a knowledge-action gap: post-trained LLMs encode signals predictive of IPI exposure, yet do not reliably bind these signals to safe agentic actions. We therefore introduce a probe-gated reasoning-based defense to bridge this gap at test time. On difficult AgentDojo settings, it substantially reduces attack success rate, e.g., from 34.6% to 0% on Qwen3.5-27B, and better preserves clean-task utility than the baselines. (3) Explanation: We introduce an analysis framework that identifies natural-language explanations strongly correlated with probe-captured signals. The resulting profiles differ across models: latent signals can align with either direct IPI-exposure sensing or indirect operational cues. Code is available at https://github.com/jianshuod/IPI-exposure-signal.

cs.CR↗

arXiv2026

Version Control System for Data with MatrixOne

The rapid advancement of artificial intelligence has elevated data to a cornerstone of modern software systems. As data projects become increasingly complex and dynamic, version control for data has become essential rather than merely convenient. Existing version control systems designed for source code are inadequate for large-scale data management, as they often require loading entire datasets into memory for diff and merge operations. Database systems, while providing robust data management capabilities, lack native support for version control operations such as diff and merge between data forks. We present a version control system for data implemented in MatrixOne, a cloud-native relational database system. Our system leverages MatrixOne's immutable storage architecture and multi-version concurrency control (MVCC) to enable git-like operations on database tables at scale. The system supports the complete spectrum of version control operations: clone, tag/branch, diff, merge, and revert, on terabyte-scale datasets with near-instantaneous performance. This version control system enables data engineers to adopt established software engineering workflows: creating branches for isolated experimentation, submitting pull requests for change review, and running CI/CD pipelines efficiently and safely. Changes in the development environment are isolated from production in both data integrity and computing resources. Verified changes can be published to production in atomic transactions, ensuring data consistency and avoiding service disruptions.

cs.DB↗

arXiv2026

Finite Difference Method for Stochastic Cahn-Hilliard Equation Driven by A Fractional Brownian Sheet

The stochastic Cahn-Hilliard equation driven by a fractional Brownian sheet provides a more accurate model for correlated space-time random perturbations. This study delves into two key aspects: first, it rigorously examines the regularity of the mild solution to the stochastic Cahn-Hilliard equation, shedding light on the intricate behavior of solutions under such complex perturbations. Second, it introduces a fully discrete numerical scheme designed to solve the equation effectively. This scheme integrates the finite difference method for spatial discretization with the tamed exponential Euler method for temporal discretization. The analysis demonstrates that the proposed scheme achieves a strong convergence rate of $O\big(h^{1-\epsilon}+\tau^{H_1-\frac{1}{8}-\frac{\epsilon}{2}}\big)$, where $\epsilon$ is an arbitrarily small positive constant, providing a solid foundation for the numerical treatment of such equations.

math.NA↗

arXiv2025

Ergodicity and invariant measure approximation of the stochastic Cahn-Hilliard equation via an explicit fully discrete scheme

This paper investigates the stochastic Cahn-Hilliard equation (SCHE) driven by additive space-time white noise. We first refine the analytical ergodic theory by proving that the continuum equation admits a unique invariant measure in the more regular state space H_\alpha, extending the classical result of Da Prato and Debussche (1996) on the negative Sobolev space $\dot{H}^{-1}_\alpha$. To approximate long-time behaviour, we introduce an explicit fully discrete scheme that combines a finite-difference spatial discretization with a strongly tamed exponential Euler method in time. Uniform-in-time moment bounds in the $L^\infty$-norm are established for the numerical solution, and a uniform strong convergence estimate with an explicit rate is derived for the fully discrete approximation. Exploiting a mass-preserving minorization tailored to Neumann boundary conditions, we further show that the numerical scheme is geometrically ergodic and possesses a unique invariant measure, together with polynomial-order error bounds for approximating the exact invariant measure. Strong laws of large numbers are proved for both the continuous and discrete systems, ensuring almost-sure convergence of temporal averages to the corresponding ergodic limits. Numerical experiments corroborate the theoretical findings, including the long-time strong convergence and the accuracy of invariant measure approximation. Overall, the results provide a complete analytical and numerical framework for investigating the long-time statistical behaviour of the SCHE.

math.NA↗

arXiv2025

PANDA: Noise-Resilient Antagonist Identification in Production Datacenters

Modern warehouse-scale datacenters commonly collocate multiple jobs on shared machines to improve resource utilization. However, such collocation often leads to performance interference caused by antagonistic jobs that overconsume shared resources. Existing antagonist-detection approaches either rely on offline profiling, which is costly and unscalable, or use a sample-from-production approach, which suffers from noisy measurements and fails under multi-victim scenarios. We present PANDA, a noise-resilient antagonist identification framework for production-scale datacenters. Like prior correlation-based methods, PANDA uses cycles per instruction (CPI) as its performance metric, but it differs by (i) leveraging global historical knowledge across all machines to suppress sampling noise and (ii) introducing a machine-level CPI metric that captures shared-resource contention among multiple co-located tasks. Evaluation on a recent Google production trace shows that PANDA ranks true antagonists far more accurately than prior methods -- improving average suspicion percentile from 50-55% to 82.6% -- and achieves consistent antagonist identification under multi-victim scenarios, all with negligible runtime overhead.

cs.PF↗

arXiv2024

Orbital cluster-based network modelling

We propose a novel reduced-order methodology to describe complex multi-frequency fluid dynamics from time-resolved snapshot data. Starting point is the Cluster-based Network Model (CNM) thanks to its fully automatable development and human interpretability. Our key innovation is to model the transitions from cluster to cluster much more accurately by replacing snapshot states with short-term trajectories ("orbits") over multiple clusters, thus avoiding nonphysical intra-cluster diffusion in the dynamic reconstruction. The proposed orbital CNM (oCNM) employs functional clustering to coarse-grain the short-term trajectories. Specifically, different filtering techniques, resulting in different temporal basis expansions, demonstrate the versatility and capability of the oCNM to adapt to diverse flow phenomena. The oCNM is illustrated on the Stuart-Landau oscillator and its post-transient solution with time-varying parameters to test its ability to capture the amplitude selection mechanism and multi-frequency behaviours. Then, the oCNM is applied to the fluidic pinball across varying flow regimes at different Reynolds numbers, including the periodic, quasi-periodic, and chaotic dynamics. This orbital-focused perspective enhances the understanding of complex temporal behaviours by incorporating high-frequency behaviour into the kinematics of short-time trajectories while modelling the dynamics of the lower frequencies. In analogy to Spectral Proper Orthogonal Decomposition, which marked the transition from spatial-only modes to spatio-temporal ones, this work advances from analysing temporal local states to examining piecewise short-term trajectories, or orbits. By merging advanced analytical methods, such as the functional representation of short-time trajectories with CNM, this study paves the way for new approaches to dissect the complex dynamics characterising turbulent systems.

physics.flu-dyn↗

arXiv2023

Dynamics-augmented cluster-based network model

In this study, we propose a novel data-driven reduced-order model for complex dynamics, including nonlinear, multi-attractor, multi-frequency, and multiscale behaviours. The starting point is a fully automatable cluster-based network model (CNM) (Li et al. J. Fluid Mech. vol.906, 2021, A21) which kinematically coarse-grains the state with clusters and dynamically predicts the transitions in a network model. In the proposed dynamics-augmented CNM (dCNM), the prediction error is reduced with trajectory-based clustering using the same number of centroids. The dCNM is first exemplified for the Lorenz system and then implemented for the three-dimensional sphere wake featuring periodic, quasi-periodic and chaotic flow regimes. For both plants, the dCNM significantly outperforms the CNM in resolving the multi-frequency and multiscale dynamics. This increased prediction accuracy is obtained by stratification of the state space aligned with the direction of the trajectories. Thus, the dCNM has numerous potential applications to a large spectrum of shear flows, even for complex dynamics.

physics.flu-dyn↗

arXiv2023

Revisit the intrinsic features of flip-flopping flow behind side-by-side circular cylinders

As one of the most intriguing wake patterns of two side-by-side circular cylinders at an intermediate gap spacing, the flip-flopping (FF) flow has attracted great attention of fundamental research interest. This FF flow is featured by the intermittently and randomly switching gap flow with correspondingly changing forces of the two cylinders. In this paper, we first present a partition map of the wake patterns behind two side-by-side circular cylinders and briefly introduce intrinsic features of each flow pattern. We focus on the FF flow aiming to explain: (i) the origin of the FF flow between laminar and turbulent regimes, (ii) their connections in different flow regimes, and (iii) mechanisms of the significantly varying flip-over time scale of the FF flows. In the laminar regime, we further divide the FF flow into the sub-classed I (FF1) and II (FF2), based on their different origins from the in-phase and anti-phase synchronized vortex shedding instabilities, respectively. By exploring the vortex interactions, we show that the FF flow in the turbulent regime has the same origin and similar vortex dynamics as the FF2 wake in the laminar regime, despite some minor disparities. Thus, a connection is established between the FF2 pattern in the laminar flow and the FF pattern in the turbulent flow. For the FF flow in the laminar regime (Re < 150-200), the mildly decreasing switching time, is several vortex shedding periods. However, for the FF flow in the weak turbulence regime (150-200 < Re < 1000-1700), the switching time scale increases significantly with Re owing to the increased vortex formation length. The FF in the strong turbulence regime (Re > 1000-1700) has a switching time scale of several orders of magnitude longer than the vortex shedding period, where the switching scale decreases gradually with Re due to the stronger Kelvin-Helmholtz vortices.

physics.flu-dyn↗

arXiv2022

Trajectory-optimized cluster-based network model for the sphere wake

We propose a novel trajectory-optimized Cluster-based Network Model (tCNM) for nonlinear model order reduction from time-resolved data following Li et al. ["Cluster-based network model, " J. Fluid Mech. 906, A21 (2021)] and improving the accuracy for a given number of centroids. The starting point is k-means++ clustering which minimizes the representation error of the snapshots by their closest centroids. The dynamics is presented by 'flights' between the centroids. The proposed trajectory-optimized clustering aims to reduce the kinematic representation error further by shifting the centroids closer to the snapshot trajectory and refining state propagation with trajectory support points. Thus, curved trajectories are better resolved. The resulting tCNM is demonstrated for the sphere wake for three flow regimes, including the periodic, quasi-periodic, and chaotic dynamics. The representation error of tCNM is 5 times smaller as compared to the approximation by the closest centroid. Thus, the error at the same level as Proper Orthogonal Decomposition (POD) of same order. Yet, tCNM has distinct advantages over POD modeling: it is human interpretable by representing dynamics by a handful of coherent structures and their transitions; it shows robust dynamics by design, i.e., stable long-time behavior; and its development is fully automatable, i.e., it does not require tuneable auxiliary closure and other models.

physics.flu-dyn↗

arXiv2022

From snapshots to manifolds - A tale of shear flows

We propose a novel non-linear manifold learning from snapshot data and demonstrate its superiority over Proper Orthogonal Decomposition (POD) for shedding-dominated shear flows. Key enablers are isometric feature mapping, Isomap (Tenenbaum et al., 2000), as encoder and K-nearest neighbours (KNN) algorithm as decoder. The proposed technique is applied to numerical and experimental datasets including the fluidic pinball, a swirling jet, and the wake behind a couple of tandem cylinders. Analyzing the fluidic pinball, the manifold is able to describe the pitchfork bifurcation and the chaotic regime with only three feature coordinates. These coordinates are linked to vortex-shedding phases and the force coefficients. The manifold coordinates of the swirling jet are comparable to the POD mode amplitudes, yet allow for a more distinct manifold identification which is less sensitive to measurement noise. As similar observation is made for the wake of two tandem cylinders (Raiola et al., 2016). The tandem cylinders are aligned in streamwise distance which corresponds to the transition between the single bluff body and the reattachment regimes of vortex shedding. Isomap unveils these two shedding regimes while the Lissajous plots of first two POD mode amplitudes feature a single circle. The reconstruction error of the manifold model is small compared to the fluctuation level, indicating that the low embedding dimensions contains the coherent structure dynamics. The proposed Isomap-KNN manifold learner is expected to be of large importance in estimation, dynamic modeling and control for large range of configurations with dominant coherent structures.

physics.flu-dyn↗

arXiv2021

Cluster-based hierarchical network model of the fluidic pinball -- Cartographing transient and post-transient, multi-frequency, multi-attractor behaviour

We propose a self-supervised cluster-based hierarchical reduced-order modelling methodology to model and analyse the complex dynamics arising from a sequence of bifurcations for a two-dimensional incompressible flow of the unforced fluidic pinball. The hierarchy is guided by a triple decomposition separating a slowly varying base flow, dominant shedding and secondary flow structures. All these flow components are kinematically resolved by a hierarchy of clusters, starting with the base flow in the first layer, resolving the vortex shedding in the second layer and distilling the secondary flow structures in the third layer. The transition dynamics between these clusters is described by a directed network, called the cluster-based hierarchical network model (HiCNM) in the sequel. Three consecutive Reynolds number regimes for different dynamics are considered: (i) periodic shedding at $Re=80$, (ii) quasi-periodic shedding at $Re=105$, and (iii) chaotic shedding at $Re=130$, involving three unstable fixed points, three limit cycles, two quasi-periodic attractors and a chaotic attractor. The HiCNM enables identifying the transient and post-transient dynamics between multiple invariant sets in a self-supervised manner. Both the global trends and the local structures during the transition are well resolved by a moderate number of hierarchical clusters. The proposed reduced-order modelling provides a visual representation of transient and post-transient, multi-frequency, multi-attractor behaviour and may automate the identification and analysis of complex dynamics with multiple scales and multiple invariant sets.

physics.flu-dyn↗

arXiv2021

The Case for Task Sampling based Learning for Cluster Job Scheduling

The ability to accurately estimate job runtime properties allows a scheduler to effectively schedule jobs. State-of-the-art online cluster job schedulers use history-based learning, which uses past job execution information to estimate the runtime properties of newly arrived jobs. However, with fast-paced development in cluster technology (in both hardware and software) and changing user inputs, job runtime properties can change over time, which lead to inaccurate predictions. In this paper, we explore the potential and limitation of real-time learning of job runtime properties, by proactively sampling and scheduling a small fraction of the tasks of each job. Such a task-sampling-based approach exploits the similarity among runtime properties of the tasks of the same job and is inherently immune to changing job behavior. Our study focuses on two key questions in comparing task-sampling-based learning (learning in space) and history-based learning (learning in time): (1) Can learning in space be more accurate than learning in time? (2) If so, can delaying scheduling the remaining tasks of a job till the completion of sampled tasks be more than compensated by the improved accuracy and result in improved job performance? Our analytical and experimental analysis of 3 production traces with different skew and job distribution shows that learning in space can be substantially more accurate. Our simulation and testbed evaluation on Azure of the two learning approaches anchored in a generic job scheduler using 3 production cluster job traces shows that despite its online overhead, learning in space reduces the average Job Completion Time (JCT) by 1.28x, 1.56x, and 1.32x compared to the prior-art history-based predictor.

cs.DC↗

arXiv2021

Dynamiques transitoires de sillage dans le pinball fluidique

In this work, we are interested in the transient dynamics of a fluid configuration consisting of three fixed cylinders whose axes distribute over an equilateral triangle in transverse flow << fluidic pinball >>. As the Reynolds number is increased on the route to chaos, its transient dynamics tell us about the contribution of the elementary degrees of freedom of the system to the lift and drag coefficients.

physics.flu-dyn↗

arXiv2021

Low-dimensional Flow Models from high-dimensional Flow data with Machine Learning and First Principles

Reduced-order modelling and system identification can help us figure out the elementary degrees of freedom and the underlying mechanisms from the high-dimensional and nonlinear dynamics of fluid flow. Machine learning has brought new opportunities to these two processes and is revolutionising traditional methods. We show a framework to obtain a sparse human-interpretable model from complex high-dimensional data using machine learning and first principles.

physics.flu-dyn↗

arXiv2021

Reduced-order modeling of the fluidic pinball

The fluidic pinball is a geometrically simple flow configuration with three rotating cylinders on the vertex of an equilateral triangle. Yet, it remains physically rich enough to host a range of interacting frequencies and to allow testing of control laws within minutes on a laptop. The system has multiple inputs (the three cylinders can independently rotate around their axis) and multiple outputs (downstream velocity sensors). Investigating the natural flow dynamics, we found that the first unsteady transition undergone by the wake flow, when increasing the Reynolds number, is a Hopf bifurcation leading to the usual time-periodic vortex shedding phenomenon, typical of cylinder wake flows, in which the mean flow field preserves axial symmetry. We extract dynamically consistent modes from the flow data in order to built a reduced-order model (ROM) of this flow regime. We show that the main dynamical features of the primary Hopf bifurcation can be described by a non-trivial lowest-order model made of three degrees of freedom.

physics.flu-dyn↗

arXiv2021

Low-order model for successive bifurcations of the fluidic pinball

We propose the first least-order Galerkin model of an incompressible flow undergoing two successive supercritical bifurcations of Hopf and pitchfork type. A key enabler is a mean-field consideration exploiting the symmetry of the mean flow and the asymmetry of the fluctuation. These symmetries generalize mean-field theory, e.g. no assumption of slow growth-rate is needed. The resulting 5-dimensional Galerkin model successfully describes the phenomenogram of the fluidic pinball, a two-dimensional wake flow around a cluster of three equidistantly spaced cylinders. The corresponding transition scenario is shown to undergo two successive supercritical bifurcations, namely a Hopf and a pitchfork bifurcations on the way to chaos. The generalized mean-field Galerkin methodology may be employed to describe other transition scenarios.

physics.flu-dyn↗