Searcharxiv⌕ Search

arXiv subjects

Zhe Zhang

Publications and source records attributed to Zhe Zhang.

At least 37 records · Page 2Linked to original sources
arXiv2026

Combining Self-Embedding Audio Watermarking with Ultra-Low-Bitrate Neural Codecs

Partial manipulation of speech recordings, where only localized segments of an utterance are altered, poses a significant challenge for content integrity verification, as reliable detection and localization of such edits becomes harder as the manipulated proportion decreases. Watermarking offers a proactive defense alternative by embedding auxiliary information prior to distribution; classical hash-based schemes achieve near-perfect detection and localization under ideal conditions, but the original content cannot be recovered once a segment is manipulated. Building on a prior self-embedding audio steganography framework, this work presents an initial exploration of proactive defense performance under ideal conditions, extending the investigation along three axes: frame-level localization, multi-bit least significant bit variants, and evaluation across multiple ultra-low-bitrate neural codec representations. By embedding a compact neural codec representation rather than a cryptographic hash, the framework additionally enables recovery of the manipulated regions, while supporting training-free detection and localization without spoofed examples. Experiments across four controlled manipulation types under ideal channel conditions show that the embedded payload, and hence an approximate reconstruction of the authentic content, is always fully recovered without bit errors. The results also indicate that the choice of neural codec is the dominant factor for detection and localization performance.

cs.SD↗
arXiv2026

Solving Convex Smooth Function Constrained Optimization Is Almost As Easy As Unconstrained Optimization

While Nesterov's Accelerated Gradient Descent (AGD) efficiently solves constrained problems when the constraint set $X \subseteq \bbr^n$ is simple and easy to project onto, it remains an open question whether function-constrained problems $\min_{x \in X} \{F(x) : g(x) \leq 0\}$ can be solved as efficiently as unconstrained problems in terms of oracle complexity. We provide an affirmative answer by proposing the Accelerated Constrained Gradient Descent (ACGD) method, a single-loop algorithm that modifies AGD by replacing the descent step with a constrained descent step, adding only a few linear constraints to the prox mapping. ACGD achieves nearly the same oracle complexity as minimizing the optimal Lagrangian function (with the multiplier fixed at its optimal value). We establish matching lower bounds, demonstrating these complexity results are unimprovable. For large-scale problems with many constraints, we introduce ACGD-S, which replaces the computationally demanding constrained descent step with basic matrix-vector multiplications, maintaining optimal oracle and computation complexities. Together, these methods provide a nearly complete characterization of the hardness of smooth function-constrained optimization. Under the additional assumption that $X$ is bounded, we also propose adaptive versions that achieve an order-optimal oracle complexity without requiring knowledge of either the aggregate smoothness constant or the diameter of $X$; only the strong convexity modulus is required in the strongly convex case. We present encouraging numerical results demonstrating their efficiency.

math.OC↗
arXiv2026

Potential Applications of HBF in LLM Serving Systems

LLM serving is increasingly constrained by memory capacity as model weights, KV caches, and the number of served model variants continue to grow. This report examines High-Bandwidth Flash (HBF) as a capacity-oriented extension to HBM-based serving systems. We first discuss how HBF can be integrated into the GPU memory hierarchy without undermining the bandwidth expected by the compute die. We then model the system-level value of added capacity as expanded residency for read-mostly model-state objects. Under this view, HBF can improve MoE serving by enabling more expert replicas and can improve multi-model serving by reducing model loading and supporting hot-model replication. Our simulation results show that these benefits depend on preserving the HBM-resident execution path while using HBF to expand the resident set of model weights.

cs.AR↗
arXiv2026

Wan-Animate-2: Pushing the Application Boundaries of Character Animation

Character image animation remains a foundational yet challenging task in computer vision. Existing approaches can be broadly categorized into three paradigms: methods based on explicit motion representations suffer from extraction errors and identity drift; methods based on implicit motion features lose fine-grained dynamics through compression; and in-context learning approaches avoid intermediate representations but incur prohibitive computational costs. Furthermore, all current systems are designed for offline synthesis, unable to meet the real-time requirements of interactive applications such as digital avatars and live-streaming hosts. To address these limitations, we present Wan-Animate-2, an end-to-end character animation framework that directly consumes the driving video within a redesigned Diffusion Transformer. Our architecture achieves superior motion fidelity and identity preservation by eliminating intermediate motion extractors entirely. We further introduce text driven viewpoint control that decouples the output camera perspective from the driving video--a capability rarely supported by prior character animation methods that rely on explicit motion representations. Beyond generation quality, we present Wan-Animate-2-Lite, an efficient variant that reduces inference latency to real-time thresholds through a three-stage training paradigm: teacher forcing pretraining with error buffer mechanism, and Self-Forcing distillation with chunk-wise backpropagation. This enables streaming character animation for interactive applications, opening new deployment scenarios that were previously infeasible. Qualitative evaluations and user studies demonstrate that Wan-Animate-2 achieves high-fidelity animation results across diverse characters and motion patterns. To foster further research and community development, we will release the Wan-Animate-2-Base model weights to the public.

cs.CV↗
arXiv2026

BulkPR-Bench: Benchmarking Queue-Level Governance of Interacting Pull Requests

Coding-agent benchmarks increasingly cover long-horizon, end-to-end, and interactive development, but typically retain one requested outcome or a fixed change sequence. Sequential policies can process a pull-request (PR) queue one candidate at a time, but when queued PRs interact, maximizing safe delivery can require jointly deciding which changes to merge and in what order. We introduce BulkPR-Bench, an executable benchmark in which an agent must recover consequential PR relations and return a large safe subset in executable order under a rolling-release protocol. The suite contains 581 newly authored candidate PRs on frozen snapshots of 18 real repositories. Registered state-by-state repository execution, including hidden safety checks, validates the gold relation graph; an exact oracle then computes the largest safe subset. Our primary metric, Relational Delivery Score (RDS), scores safe delivery and correct rejection over relation groups from the realized merge trace; Global Safety-Gated Yield (Global-SGY) separately measures strict delivery of the realized whole-queue plan. Under the buffered primary protocol with batch size $K=32$, the three highest RDS estimates among the six models are 66.6%, 62.0%, and 57.9%, compared with 53.1% for the strongest sequential baseline. Only 8 of 324 model runs complete a queue exactly. Critical-relation recall ranges from 35.2% to 57.7%, and diagnostic runs supplied with the gold relations show substantial remaining headroom. Gains on relation groups therefore do not yet translate into dependable whole-queue governance.

cs.SE↗
arXiv2026

XMatchAD: A Cross-Modal Matching Perspective on Reconstruction-based Anomaly Detection

The remarkable success of reconstruction-based methods in Unsupervised Anomaly Detection (UAD) lies in their ability to identify and localize anomalies by modeling discrepancies between input images and their reconstructed counterparts. However, these approaches often struggle to capture subtle anomalies and tend to produce blurred anomaly boundaries, which significantly limits their effectiveness, particularly in complex multi-class scenarios. To address these issues, we present XMatchAD, a novel UAD framework that reinterprets the task from a pseudo cross-modal matching perspective. Specifically, the input and reconstructed images are treated as two complementary modalities and their matching relationships are precisely exploited for anomaly detection. First, a pre-trained feature extractor is employed to encode discriminative representations. Second, an attention-guided cross-modal matching mechanism is introduced to match local inter-modal anomaly-related patterns while mutually refining the features. This enhances the sensitivity to anomalies with diverse shapes and subtle deviations and significantly improves the precision of anomaly detection and localization. Third, we design an adaptive frequency-aware fusion module that further delineates sharp anomaly boundaries through the coupling of high-frequency components from cross-modal multi-scale representations. Comprehensive evaluations on MVTec-AD, VisA, and MPDD benchmarks demonstrate that our method consistently achieves superior performance, outperforming state-of-the-art methods in multi-class anomaly detection and localization. The code will be released at https://github.com/Mingxiu-Cai/XMatchAD.

cs.CV↗
arXiv2026

Impedance Control of Ship-Borne Manipulators via Optimization-based Task-Space Inverse Dynamics

Ship-borne manipulators operating in maritime environments are subject to stochastic wave-induced base motions that introduce kinematic disturbances and dynamic coupling, degrading trajectory tracking accuracy and complicating safe, contact-rich manipulation. This paper proposes a torque-level optimization-based control framework that integrates high-precision trajectory tracking with task-space impedance for ship-borne manipulators. The controller is formulated using task-space inverse dynamics (TSID) and solved via quadratic programming to explicitly compensate for the dynamic coupling introduced by base motion. To enable accurate feedforward compensation, an error-state Kalman filter (ESKF) is developed to estimate the base state by fusing inertial measurements with end-effector pose feedback. The framework is validated in simulation and real-world experiments using a 7-DOF manipulator mounted on a 6-DOF Stewart platform. The proposed method reduces real-world end-effector position tracking error by over 25.7% compared with the best baseline. Furthermore, the controller enables dynamic peg-in-hole insertion with 1~mm clearance under base motion, increasing the success rate while reducing average contact forces by 45%, demonstrating precise and compliant manipulation in contact-rich environments.

cs.RO↗
arXiv2026

Accuracy Analysis of VLBI Universal Time Measurement Based on a GNSS Single-Station Regional Ionospheric Model

Universal Time (UT1) is a key parameter characterizing Earth's rotation, and very long baseline interferometry (VLBI) is the mainstream technique for measuring UT1. To address the limitations in the timeliness and accuracy of existing global ionospheric models for single-frequency VLBI UT1 measurements, we construct a single-station regional ionospheric model using GNSS data from the VLBI stations on the Jilin-Kashi baseline. We apply this model to VLBI observations and compare its correction performance with that of a global predictive model and a global post-processed model. The results show that the line-of-sight ionospheric delays and baseline corrections calculated with the single-station regional model have precision close to that of the global post-processed model and are substantially better than those of the global predictive model. After correction with the single-station regional model, the derived UT1 values differ from the US Naval Observatory (USNO) reference values by a mean bias of -15.6 us and an RMS deviation of 82.3 us, both better than the results obtained with the other two model classes. A single-station regional ionospheric model constructed independently from GNSS data available at VLBI stations can effectively correct single-frequency VLBI observations and support quasi-real-time high-precision UT1 measurements. It therefore has important value for improving the timeliness of independent UT1 products.

astro-ph.EP↗
arXiv2026

Optimal Parameter-Free First-Order Methods for Convex Optimization with Unknown Growth and Smoothness

We study deterministic first-order minimization of a convex function without prior knowledge of the objective's growth, smoothness regime, or associated parameters. We develop anytime, parameter-free bundle-level methods that adapt simultaneously to these unknown properties and attain best-known oracle complexities. For nonsmooth Lipschitz objectives satisfying quadratic growth, the proposed bundle-level W-certificate method (BLW) achieves the optimal complexity without requiring the growth modulus or target accuracy as input. We then introduce an accelerated variant, A-BLW. Without knowing the Hölder smoothness parameters, the quadratic-growth modulus, or the target accuracy, A-BLW attains the optimal rates in the nonsmooth, weakly smooth, and smooth regimes. Central to both methods is an affine W-certificate, a condition based on the descent-slowness of an affine minorant that converts the geometry of a bundle model into an optimality-gap guarantee under quadratic growth. A stopping-time analysis further shows that the same A-BLW algorithm, without modification, achieves the corresponding best-known rates for general convex objectives and for objectives satisfying Hölder growth of order at least two. Numerical experiments illustrate the practical performance of the proposed methods.

math.OC↗
arXiv2026

Nearly Optimal $L_p$ Risk Minimization

Convex risk measures play a foundational role in the area of stochastic optimization. However, in contrast to risk neutral models, their applications are still limited due to the lack of efficient solution methods. In particular, the mean $L_p$ semi-deviation is a classic risk minimization model, but its solution is highly challenging due to the composition of concave-convex functions and the lack of uniform Lipschitz continuity. In this paper, we discuss some progresses on the design of efficient algorithms for $L_p$ risk minimization, including a novel lifting reformulation to handle the concave-convex composition, and a new stochastic approximation method to handle the non-Lipschitz continuity. We establish an upper bound on the sample complexity associated with this approach and show that this bound is not improvable for $L_p$ risk minimization in general through the construction of a nearly matching lower complexity bound.

math.OC↗
arXiv2026

Distance Determination of Southern Galactic Plane Supernova Remnants with the Mopra CO Survey and DECaPS 3D Dust Map

Accurate distance measurements to supernova remnants (SNRs) are crucial for understanding their physical properties, evolutionary processes, and role in the Galactic interstellar medium (ISM) cycle. In this study, we apply for the first time to the southern Galactic plane a distance determination method that utilizes CO emission data from the Mopra survey to identify molecular clouds (MCs) interacting with SNRs. By combining this with extinction-distance profiles from the DECaPS three-dimensional (3D) extinction map, we directly measure the distances to the associated MCs, thereby obtaining precise distances to the remnants. To overcome the extinction-missing bias in extremely dense regions where the 3D map suffers from a deficit of background stars, we supplement our analysis with two-dimensional (2D) extinction maps as cross-validation. Applying this method, we have derived precise distances for nine SNRs: G290.1-0.8 (7.32+0.60/-0.47 kpc), G292.2-0.5 (10.85+0.43/-0.68 kpc), G296.1-0.5 (4.59+0.18/-0.19 kpc), G296.8-0.3 (8.74+0.40/-0.29 kpc), G298.6-0.0 (6.50 +/- 0.21 kpc), G312.4-0.4 (3.60+0.19/-0.23 kpc), G332.4-0.4 (2.66+0.23/-0.15 kpc), G335.2+0.1 (2.76+0.37/-0.31 kpc), and G353.6-0.7 (1.81+0.18/-0.14 kpc). Additionally, we established a robust lower distance limit of 1.34 kpc for G351.7+0.8.

astro-ph.GA↗
arXiv2026

Synthetic Audio Generation Framework for Air Traffic Control Speech Recognition

Automatic Speech Recognition (ASR) systems, despite achieving remarkable accuracy in general-purpose domains with native speech (L1), struggle in domains like Air Traffic Control (ATC) due to strong channel noise, a presence of non-native (L2) English accents, and data scarcity. We propose a synthetic data generation pipeline with acoustical properties simulations specifically designed to address this lack of real data to improve recognition accuracy in the ATC domain. Our approach leverages a combination of neural generation techniques, including Text-to-Speech, Voice Conversion, L2-to-L1 accent conversion, and a novel controllable L1-to-L2 accent conversion framework built to simulate accented speech. Our experiments with the Whisper model on the ATCO2 corpus demonstrate that fine-tuning with either synthetic data alone, or a mix of real and synthetic data, significantly improves the word error rate over out-of-the-box and real data only baselines respectively.

cs.CL↗
arXiv2026

ARSAR-Net: Adaptively Regularized SAR Imaging Network with Efficient Unfolding

Developed from sparse reconstruction approaches, deep unfolding networks (DUNs) have constituted an emerging method for synthetic aperture radar (SAR) imaging, offering fast convergence and data-driven learning. However, baseline unfolding networks, derived from iterative sparse reconstruction algorithms such as alternating direction method of multipliers (ADMM), lack generalization capability across scenes, as their regularizers are empirically designed and keep unchanged during imaging. In this study, we introduce a learnable regularizer to the unfolding network and propose an adaptively regularized SAR imaging network (ARSAR-Net) for scene-agnostic imaging (imaging across heterogeneous scenes of varying sparsity levels). In practice, the vanilla ARSAR-Net suffers from inherent structural limitations in 2D signal processing, primarily due to its reliance on matrix inversion. To conquer this, we further develop an ADMM without matrix inversion for efficient unfolding, by designing linear operations to replace the time-consuming matrix inversion operations. Experiments upon simulated and real-data demonstrate three advantages of ARSAR-Net: (1) a PSNR gain of up to 2.0 dB in imaging quality compared to existing deep network based imaging methods, (2) enhanced adaptability to complex scenes, and (3) a 50\% increase in imaging speed over existing unfolding networks. These advancements establish a new paradigm for efficient and scene-agnostic SAR imaging systems. Code is available at github.com/ShipenFyu/ARSAR-Net.

eess.SP↗
arXiv2026

SAGE: Scalable AI Governance & Evaluation

Evaluating relevance in large-scale search systems is fundamentally constrained by the governance gap between nuanced, resource-constrained human oversight and the high-throughput requirements of production systems. While traditional approaches rely on engagement proxies or sparse manual review, these methods often fail to capture the full scope of high-impact relevance failures. We present \textbf{SAGE} (Scalable AI Governance \& Evaluation), a framework that operationalizes high-quality human product judgment as a scalable evaluation signal. At the core of SAGE is a bidirectional calibration loop where natural-language \emph{Policy}, curated \emph{Precedent}, and an \emph{LLM Surrogate Judge} co-evolve. SAGE systematically resolves semantic ambiguities and misalignments, transforming subjective relevance judgment into an executable, multi-dimensional rubric with near human-level agreement. To bridge the gap between frontier model reasoning and industrial-scale inference, we apply teacher-student distillation to transfer high-fidelity judgments into compact student surrogates at \textbf{92$\times$} lower cost. Deployed within LinkedIn Search ecosystems, SAGE guided model iteration through simulation-driven development, distilling policy-aligned models for online serving and enabling rapid offline evaluation. In production, it powered policy oversight that measured ramped model variants and detected regressions invisible to engagement metrics. Collectively, these drove a \textbf{0.25\%} lift in LinkedIn daily active users.

cs.IR↗
arXiv2026

Accelerated Prox-Level Methods for Unknown Piecewise-Smooth Optimization I: Convex Optimization

We introduce a nearly parameter-free algorithm for minimizing piecewise smooth (PWS) convex functions under the quadratic-growth (QG) condition, where the locations and structure of the smooth regions are entirely unknown. Our algorithm, APEX (Accelerated Prox-Level method for Exploring Piecewise Smoothness), is an accelerated bundle-level method designed to adaptively exploit the underlying PWS structure. For this setting, APEX achieves the best-known oracle-complexity result among existing first-order methods, improving the dependence on the condition number relative to prior bundle-level guarantees. Furthermore, APEX generates a verifiable and accurate termination certificate, enabling a robust, nearly parameter-free implementation. To the best of our knowledge, APEX is the first algorithm to simultaneously achieve the best-known first-order oracle complexity for PWS optimization and provide certificate guarantees.

math.OC↗
arXiv2026

Simultaneous Contact Selection and Planning for Contact-Rich Manipulation with Cascaded Optimization

We propose an optimization-based framework for robust contact-rich manipulation. Recent contact-implicit methods enable online hybrid planning across contact modes, allowing closed-loop manipulation for a given target state and contact location sequence of the robot and object. However, most existing approaches lack the ability to autonomously reason and generate diverse contact location sequences and manipulation trajectories, i.e., active contact location selection, which limits their applicability to relatively simple tasks. Active contact location selection is challenging due to complementarity in contact dynamics and the sparse gradients, making the design of a unified framework for contact selection and planning difficult. To address these challenges, we introduce Simultaneous Contact Selection and Planning (SCSP), a cascaded optimization framework comprising Contact Selection Optimization (CSO) and Contact Planning Optimization (CPO). CSO leverages a surrogate contact model and discrete-continuous optimization to efficiently resolve the nonsmoothness and coupling in contact selection, enabling online global searching of optimal contact locations. CPO performs prior-guided contact planning by evaluating the reference contact locations produced by CSO and generating corresponding manipulation trajectories in real time for redundant manipulators. Extensive simulations and real-world experiments demonstrate that SCSP produces diverse manipulation behaviors and robust control under inaccurate dynamics and perceptual noise. We further validate the generalization of the framework on challenging manipulation tasks. Project website: \href{https://sites.google.com/view/scsp-robot}{https://sites.google.com/view/scsp-robot}.

cs.RO↗
arXiv2026

Decoupling Variance and Scale-Invariant Updates in Adaptive Gradient Descent for Unified Vector and Matrix Optimization

Adaptive methods like Adam have become the $\textit{de facto}$ standard for large-scale vector and Euclidean optimization due to their coordinate-wise adaptation with a second-order nature. More recently, matrix-based spectral optimizers like Muon (Jordan et al., 2024b) show the power of treating weight matrices as matrices rather than long vectors. Linking these is hard because many natural generalizations are not feasible to implement, and we also cannot simply move the Adam adaptation to the matrix spectrum. To address this, we reformulate the AdaGrad update and decompose it into a variance adaptation term and a scale-invariant term. This decoupling produces $\textbf{DeVA}$ ($\textbf{De}$coupled $\textbf{V}$ariance $\textbf{A}$daptation), a framework that bridges between vector-based variance adaptation and matrix spectral optimization, enabling a seamless transition from Adam to adaptive spectral descent. Extensive experiments across language modeling and image classification demonstrate that DeVA consistently outperforms state-of-the-art methods such as Muon and SOAP (Vyas et al., 2024), reducing token usage by around 6.6\%. Theoretically, we show that the variance adaptation term effectively improves the blockwise smoothness, facilitating faster convergence. Our implementation is available at https://github.com/Tsedao/Decoupled-Variance-Adaptation

cs.LG↗
arXiv2026

Can Entry-Wise Clipping Give Spectral Control of Stochastic Gradients?

Training instabilities such as loss spikes are frequently the result of stochastic gradient noise. Because of rare expressions in language training data, and multiple layer composition, the noise impact is heavy-tailed and survives mini-batch averaging. Existing remedies trade off structure against cost: vector-norm clipping ignores the matrix structure of weight updates, while spectral normalization (e.g., Muon (Jordan et al., 2024)) respects it at additional cost. We show that this trade-off can be balanced. Real gradient noise appears to be similar to entry-wise heavy-tailed contamination, and a first-order perturbation analysis reveals a localization property of such noise, under which a simple entry-wise method achieves spectral control. Exploiting this, we derive a tractable surrogate for the Bayes-optimal entry-wise estimator under a Gaussian signal prior. We establish $O(ε^{-4})$ convergence guarantee under Cauchy-contaminated noise. Empirically, we find that smooth shrinkage improves Adam on NanoGPT pretraining, saving ${\sim}7\%$ of training tokens. We further find that applying the entry-wise clipping before spectral normalization yields a ${\sim}2\%$ token saving on top of Muon.

cs.LG↗