arXiv subjects

Weiye Zhang

Publications and source records attributed to Weiye Zhang.

4 recordsLinked to original sources

arXiv2026

CF-VLA: Efficient Coarse-to-Fine Action Generation for Vision-Language-Action Policies

Flow-based vision-language-action (VLA) policies offer strong expressivity for action generation, but suffer from a fundamental inefficiency: multi-step inference is required to recover action structure from uninformative Gaussian noise, leading to a poor efficiency-quality trade-off under real-time constraints. We address this issue by rethinking the role of the starting point in generative action modeling. Instead of shortening the sampling trajectory, we propose CF-VLA, a coarse-to-fine two-stage formulation that restructures action generation into a coarse initialization step that constructs an action-aware starting point, followed by a single-step local refinement that corrects residual errors. Concretely, the coarse stage learns a conditional posterior over endpoint velocity to transform Gaussian noise into a structured initialization, while the fine stage performs a fixed-time refinement from this initialization. To stabilize training, we introduce a stepwise strategy that first learns a controlled coarse predictor and then performs joint optimization. Experiments on CALVIN and LIBERO show that our method establishes a strong efficiency-performance frontier under low-NFE (Number of Function Evaluations) regimes: it consistently outperforms existing NFE=2 methods, matches or surpasses the NFE=10 $\pi_{0.5}$ baseline on several metrics, reduces action sampling latency by 75.4%, and achieves the best average real-robot success rate of 83.0%, outperforming MIP by 19.5 points and $\pi_{0.5}$ by 4.0 points. These results suggest that structured, coarse-to-fine generation enables both strong performance and efficient inference. Our code is available at https://github.com/EmbodiedAI-RoboTron/CF-VLA.

cs.CV↗

arXiv2025

Generalized probabilistic approximation characteristic based on Birkhoff orthogonality and related conclusions in $S_{\infty}$-norm

In this article, we generalize the definition of the probabilistic Gel'fand width from the Hilbert space to the strictly convex reflexive space by giving Birkhoff left orthogonal decomposition theorem. Meanwhile, a more natural definition of Gel'fand width in the classical setting is selected to make sure probabilistic and average Gel'fand widths will not lose their meaning, so that we can give the equality relation between the probabilistic Gel'fand width and the probabilistic linear width of the Hilbert space. Meanwhile, we use this relationship to continue the study of the Gel'fand widths of the univariate Sobolev space and the multivariate Sobolev space, especially in $S_{\infty}$-norm, and determine the exact order of probabilistic and average Gel'fand widths.

math.FA↗

arXiv2024

ETGuard: Malicious Encrypted Traffic Detection in Blockchain-based Power Grid Systems

The escalating prevalence of encryption protocols has led to a concomitant surge in the number of malicious attacks that hide in encrypted traffic. Power grid systems, as fundamental infrastructure, are becoming prime targets for such attacks. Conventional methods for detecting malicious encrypted packets typically use a static pre-trained model. We observe that these methods are not well-suited for blockchain-based power grid systems. More critically, they fall short in dynamic environments where new types of encrypted attacks continuously emerge. Motivated by this, in this paper we try to tackle these challenges from two aspects: (1) We present a novel framework that is able to automatically detect malicious encrypted traffic in blockchain-based power grid systems and incrementally learn from new malicious traffic. (2) We mathematically derive incremental learning losses to resist the forgetting of old attack patterns while ensuring the model is capable of handling new encrypted attack patterns. Empirically, our method achieves state-of-the-art performance on three different benchmark datasets. We also constructed the first malicious encrypted traffic dataset for blockchain-based power grid scenario. Our code and dataset are available at https://github.com/PPPmzt/ETGuard, hoping to inspire future research.

cs.CR↗

arXiv2023

GO-FEAP: Global Optimal UAV Planner Using Frontier-Omission-Aware Exploration and Altitude-Stratified Planning

Autonomous exploration is a fundamental problem for various applications of unmanned aerial vehicles(UAVs). Existing methods, however, are demonstrated to static local optima and two-dimensional exploration. To address these challenges, this paper introduces GO-FEAP (Global Optimal UAV Planner Using Frontier-Omission-Aware Exploration and Altitude-Stratified Planning), aiming to achieve efficient and complete three-dimensional exploration. Frontier-Omission-Aware Exploration module presented in this work takes into account multiple pivotal factors, encompassing frontier distance, nearby frontier count, frontier duration, and frontier categorization, for a comprehensive assessment of frontier importance. Furthermore, to tackle scenarios with substantial vertical variations, we introduce the Altitude-Stratified Planning strategy, which stratifies the three-dimensional space based on altitude, conducting global-local planning for each stratum. The objective of global planning is to identify the most optimal frontier for exploration, followed by viewpoint selection and local path optimization based on frontier type, ultimately generating dynamically feasible three-dimensional spatial exploration trajectories. We present extensive benchmark and real-world tests, in which our method completes the exploration tasks with unprecedented completeness compared to state-of-the-art approaches.

cs.RO↗