arXiv subjects

Long Huang

Publications and source records attributed to Long Huang.

At least 19 recordsLinked to original sources

arXiv2026

The Information Rate of Fiber-Wireless Communication Systems Based on Photonic Generation of RF Signals

High-capacity fiber-wireless communication systems operating at high frequencies increasingly rely on photonic generation of radio-frequency (RF) signals. In these systems, optical signals are transmitted over optical fibers and detected by photodetectors, where RF signals are generated at frequencies equal to the difference between the optical carrier frequencies. A major performance-limiting impairment is the phase noise of the generated RF signals, which originates from the phase noise of the optical sources. In this paper, we develop comprehensive probabilistic models for the two principal configurations of fiber-wireless communication systems employing photonic RF generation. Based on these models, we propose an efficient numerical framework for calculating the information rate (IR). Numerical simulations are performed to validate the efficiency of the proposed algorithm and the results provide design guidelines for high-performance fiber-wireless systems.

physics.optics↗

arXiv2026

The distance from functions in BMO to BLO

Let BMO and BLO denote the spaces of all locally integrable real-valued functions on $\mathbb{R}^n$ with bounded mean oscillation and bounded lower oscillation, respectively. It is well known that $$L^\infty(\mathbb{R}^n)\subsetneqq {\rm BLO}\subsetneqq {\rm BMO}.$$ In 1978, Garnett and Jones gave distance formulas of $f\in {\rm BMO}$ to $L^\infty(\mathbb{R}^n)$ and recently, Angrisani studied the distance of $f\in {\rm BLO}$ to $L^\infty(\mathbb{R}^n)$. In this paper, we characterize the distance from any given function $f \in {\rm BMO}$ to BLO via the Muckenhoupt weight class $A_p$ as follows \begin{center} dist\,($f$,\ BLO)\,$\sim\inf\left\{\xi\in(0,\infty):\ e^{-\frac f\xi}\in A_p\ \mathrm{for\ some\ }p\in(1,\infty)\right\}$. \end{center} Two equivalent representations of this distance are also established in terms of exponential form and the infimum of the constant in a variant of John--Nirenberg inequality, respectively.

math.CA↗

arXiv2026

From Independent to Joint: Enhancing Quantum Phase and Correlation Factor Estimation by Squeezed Reservoir Engineering

High-precision quantum parameter estimation is fundamental to the advancement of quantum metrology. Although reservoir engineering provides a powerful approach to improve estimation by tailoring system-environment interactions, the role of the squeezing phase and correlations arising from the sequential utilization of the same squeezed reservoir remains inadequately explored. In this work, we employ a correlated squeezed-thermal reservoir to enhance the precision of estimating the phase parameter $\phi$ and the correlation factor $\mu$, both individually and simultaneously. We show that the squeezing phase $\Phi$ is crucial for achieving quantum-enhanced precision, with optimal phase-matching conditions that depend strongly on $\mu$. Specifically, we derive the near-optimal phase-matching relations aimed at maximizing the quantum Fisher information (QFI) for both $\phi$ and $\mu$, as well as minimizing the total variance $\Delta_{\rm sim}$ in joint estimation. Furthermore, we show that the joint estimation variance is dominated by $F_{\phi}$, which motivates our search for the phase-matching conditions that minimize $\Delta_{\text{sim}}$. Through the ratio $R$ of variances, we demonstrate that joint estimation conserves quantum resources and maintains high precision when the squeezing phase is optimized for $F_{\phi}$, despite the inherent incompatibility of the parameters. These findings provide practical insights into reservoir engineering strategies for high-precision quantum sensing and information processing.

quant-ph↗

arXiv2026

Resource-Efficient Teleportation of High-Dimensional Quantum Coherence via Initial Phase Engineering

High-dimensional quantum systems leverage an expanded Hilbert space to enhance resilience against decoherence and noise. However, standard quantum teleportation is fundamentally limited by the quadratic growth of measurement complexity and high classical communication overhead, requiring the resolution of $d^2$ Bell states and $2\log_2 d$ classical bits. In this study, we propose a resource-efficient high-dimensional coherence teleportation (REHDCT) protocol. By designing $d$ sets of specialized positive operator-valued measure (POVM) bases, our protocol achieves a 50\% reduction in classical communication by utilizing one of the $d$ designed POVM sets, which effectively scales the measurement complexity from $O(d^2)$ to $O(d)$. Furthermore, we demonstrate that by utilizing initial phase engineering to align the target qudit with the measurement basis, theoretically perfect teleportation of quantum coherence can be achieved for arbitrary qudit states. A quantitative robustness analysis reveals that the protocol remains highly resilient to operational errors, maintaining an efficiency above 99.6\% even under a 0.1 rad phase deviation for $d=16$. Our analysis under various noise models (amplitude damping, phase flip, depolarizing, and dit-flip) confirms that high-dimensional systems exhibit an expanding quantum advantage window as dimensionality increases. Notably, under dit-flip noise, perfect coherence teleportation can be restored through the optimal selection of the POVM basis. These findings establish REHDCT as a practical, hardware-friendly framework for resource-efficient quantum communication in future high-dimensional networks.

quant-ph↗

arXiv2026

Uncertainty-Aware Gradient Signal-to-Noise Data Selection for Instruction Tuning

Instruction tuning is a standard paradigm for adapting large language models (LLMs), but modern instruction datasets are large, noisy, and redundant, making full-data fine-tuning costly and often unnecessary. Existing data selection methods either build expensive gradient datastores or assign static scores from a weak proxy, largely ignoring evolving uncertainty, and thus missing a key source of LLM interpretability. We propose GRADFILTERING, an objective-agnostic, uncertainty-aware data selection framework that utilizes a small GPT-2 proxy with a LoRA ensemble and aggregates per-example gradients into a Gradient Signal-to-Noise Ratio (G-SNR) utility. Our method matches or surpasses random subsets and strong baselines in most LLM-as-a-judge evaluations as well as in human assessment. Moreover, GRADFILTERING-selected subsets converge faster than competitive filters under the same compute budget, reflecting the benefit of uncertainty-aware scoring.

cs.CL↗

arXiv2025

Microcomb-driven large-scale fully connected quantum network

Fully connected quantum networks enable simultaneously connecting every user to every other user and are the most versatile and robust networking architecture. However, the scalability of such networks remains great challenge for practical applications. Here we construct a large-scale fully connected quantum network founded on two-photon Hong-Ou-Mandel (HOM) interference, where user-to-user security is guaranteed even with untrusted network provider. Using integrated soliton microcomb (SMC) and photonic encoding chips, we realize precise massive parallel frequency generation and locking, high-visibility HOM interferences and measurement-device-independent (MDI) quantum key distribution. The proposed architecture enables a 200-user fully connected quantum network over 200 kilometers with strict information-theoretic security via untrusted network provider. The implemented networking architecture paves the way for realizing large-scale fully connected MDI quantum networks across metropolitan and intercity regions.

quant-ph↗

arXiv2025

HyperVL: An Efficient and Dynamic Multimodal Large Language Model for Edge Devices

Current multimodal large lanauge models possess strong perceptual and reasoning capabilities, however high computational and memory requirements make them difficult to deploy directly on on-device environments. While small-parameter models are progressively endowed with strong general capabilities, standard Vision Transformer (ViT) encoders remain a critical bottleneck, suffering from excessive latency and memory consumption when processing high-resolution inputs.To address these challenges, we introduce HyperVL, an efficient multimodal large language model tailored for on-device inference. HyperVL adopts an image-tiling strategy to cap peak memory usage and incorporates two novel techniques: (1) a Visual Resolution Compressor (VRC) that adaptively predicts optimal encoding resolutions to eliminate redundant computation, and (2) Dual Consistency Learning (DCL), which aligns multi-scale ViT encoders within a unified framework, enabling dynamic switching between visual branches under a shared LLM. Extensive experiments demonstrate that HyperVL achieves state-of-the-art performance among models of comparable size across multiple benchmarks. Furthermore, it significantly significantly reduces latency and power consumption on real mobile devices, demonstrating its practicality for on-device multimodal inference.

cs.CV↗

arXiv2025

Capacitary Muckenhoupt Weight, BMO and BLO Spaces with Hausdorff Content, Factorization Theorems and Applications

Let $\delta\in(0,n]$, $p\in[1,\infty)$, $\mathcal H_{\infty}^\delta$ denote the Hausdorff content on $\mathbb R^n$, and $\mathcal A_{p,\delta}$ be the capacitary Muckenhoupt weight class. We are interested in understanding the relationship between the capacitary Muckenhoupt weight class $\mathcal A_{p,\delta}$ and ${\rm{BMO}}(\mathbb R^n, \mathcal H_{\infty}^{\delta})$ or ${\rm{BLO}}(\mathbb R^n, \mathcal H_{\infty}^{\delta})$ spaces for all dimension $\delta\in(0,n]$, and further to comprehend the structure of these two spaces. Our main result shows that $\mathcal A_{p,\delta}$ for $p\in(1,\infty)$ is equivalent to the BMO spaces, while $\mathcal A_{1,\delta}$ is equivalent to the BLO spaces, and consequently yields the factorization theorems for these BMO and BLO spaces via capacitary Hardy--Littlewood maximal operators, which essentially extend main results of Coifman and Rochberg in 1980 beyond measure theory. As applications, by establishing some capacitary weighted John--Nirenberg inequalities, we obtain the equivalence between capacitary weighted BMO or BLO spaces and ${\rm{BMO}}(\mathbb R^n, \mathcal H_{\infty}^{\delta})$ or ${\rm{BLO}}(\mathbb R^n, \mathcal H_{\infty}^{\delta})$ respectively. These results reveal deep connections between $\mathcal A_{p,\delta}$ and BMO or BLO spaces with Hausdorff content, beyond the classical measure-theoretic settings. We develop some approaches in the proofs and using a new observation, that is, the additivity of measures and linearity of integrals are superfluous for the corresponding classical theory.

math.CA↗

arXiv2025

Capacitary Muckenhoupt Weights and Weighted Norm Inequalities for Hardy-Littlewood Maximal Operators

Let $\mathcal H_{\infty}^\delta$ denote the Hausdorff content of dimension $\delta\in(0,n]$ defined on subsets of $\mathbb R^n$. The principal problem, considered in this paper, is to characterize the non-negative function $w$ for which the weighted $L^p$-norm inequality with $p\in(1,\infty)$ and the weighted weak $L^1$-norm inequality on Hardy-Littlewood maximal operators associated with Hausdorff contents hold true. To achieve this, we introduce a class of capacitary Muckenhoupt weights depending on the dimension $\delta$, denoted as $\mathcal A_{p,\delta}$, which enjoys the strict monotonicity on the dimension index $\delta$. Then we show that, for any $p\in(1,\infty)$ and $\delta\in(0,n]$, the weighted $L^p$-norm inequality holds true if and only if $w\in\mathcal A_{p,\delta}$, and the weighted weak $L^1$-norm inequality holds true if and only if $w\in\mathcal A_{1,\delta}$ by a new approach developed in this paper. As the second objective, applying this new approach, the seminal properties of classical Muckenhoupt $A_p$ weights, such as the reverse H\"older inequality [R. R. Coifman and C. Fefferman, Studia Math. 51 (1974), 241-250], the self-improving property [B. Muckenhoupt, Trans. Amer. Math. Soc. 165 (1972), 207-226], and Jones' factorization theorem [P. W. Jones, Ann. of Math. (2) 111 (1980), 511-530], are all established within the framework of capacitary Muckenhoupt weight class $\mathcal A_{p,\delta}$. Finally, we also show that the maximal operator is bounded on the weak weighted Choquet-Lebesgue space $L_w^{p,\infty}(\mathbb R^n,{\mathcal H}_\infty^\delta)$ if and only if $w\in\mathcal A_{p,\delta}$ with $p\in(1,\infty)$ and $\delta\in(0,n]$.

math.CA↗

arXiv2025

Taming the Chaos: Coordinated Autoscaling for Heterogeneous and Disaggregated LLM Inference

Serving Large Language Models (LLMs) is a GPU-intensive task where traditional autoscalers fall short, particularly for modern Prefill-Decode (P/D) disaggregated architectures. This architectural shift, while powerful, introduces significant operational challenges, including inefficient use of heterogeneous hardware, network bottlenecks, and critical imbalances between prefill and decode stages. We introduce HeteroScale, a coordinated autoscaling framework that addresses the core challenges of P/D disaggregated serving. HeteroScale combines a topology-aware scheduler that adapts to heterogeneous hardware and network constraints with a novel metric-driven policy derived from the first large-scale empirical study of autoscaling signals in production. By leveraging a single, robust metric to jointly scale prefill and decode pools, HeteroScale maintains architectural balance while ensuring efficient, adaptive resource management. Deployed in a massive production environment on tens of thousands of GPUs, HeteroScale has proven its effectiveness, increasing average GPU utilization by a significant 26.6 percentage points and saving hundreds of thousands of GPU-hours daily, all while upholding stringent service level objectives.

cs.DC↗

arXiv2025

Quantitative Hardy--Littlewood maximal inequalities and Wiener--Stein theorem on p.c.f. fractals

Let $K\subset \mathbb{R}^d$ be a post-critically finite (p.c.f.) self-similar set with Hausdorff dimension $s$, and $\mu$ be a self-similar probability measure supported on $K$. Let $H^{\alpha}_\mu$, $0<\alpha\le s$, be the Hausdorff content on $K$, and $M_{\mathcal{D}}^\mu $ be the Hardy--Littlewood maximal operator defined on $K$ associated with its basic cubes $\mathcal{D}$. In this paper, we establish quantitative strong type and weak type Hardy--Littlewood maximal inequalities on fractal set $K$ with respect to $H^{\alpha}_\mu$ for all range $0<\alpha\le s$. As applications, the Lebesgue differentiation theorem on $K$ is proved. Moreover, via the Hardy--Littlewood maximal operator $M_{\mathcal{D}}^\mu $, we characterize the Lebesgue--Choquet space $L^p(K,H^{\alpha}_\mu)$ and the Zygmund space $L\log L(K,\mu)$. To be exact, given $\alpha/s< p\le \infty$, we discover that \[ \text{$f\in L^p(K,H^{\alpha}_\mu)$ if and only if $M_{\mathcal{D}}^\mu f\in L^p(K,H^{\alpha}_\mu)$}\] and, for $f\in L^1(K,\mu)$ with $K$ satisfying the strong separation condition, \[\text{$M_{\mathcal{D}}^\mu f\in L^1(K,\mu)$ if and only if $f\in L\log L(K,\mu)$}.\] That is, Wiener's $L\log L$ inequality and its converse inequality due to Stein in 1969 are extended to fractal set $K$ with respect to $\mu$.

math.FA↗

arXiv2025

Silicon Micro-Disk Resonator Crossbar Array for High-Speed and High-Density Photonic Convolution Processing

Advanced artificial intelligence (AI) algorithms, particularly those based on artificial neural networks, have garnered significant attention for their potential applications in areas such as image recognition and natural language processing. Notably, neural networks make heavy use of matrix-vector multiplication (MVM) operations, causing substantial computing burden on existing electronic computing systems. Optical computing has attracted considerable attention that can perform optical-domain MVM at an ultra-high speed. In this paper, we introduce a novel silicon photonic micro-disk resonator (MDR) crossbar signal processor designed to support matrix-vector multiplication (MVM) with both high processing speed and enhanced computational density. The key innovation of the proposed MDR crossbar processor is the placement of two MDRs at each crosspoint, enabling simultaneous routing and weighting functions. This design effectively doubles the computational density, improving overall performance. We fabricate a silicon photonic MDR crossbar processor, which is employed to perform convolutional tasks in a convolutional neural network (CNN). The experimental results demonstrate that the photonic processor achieves a classification accuracy of 96% on the MNIST dataset. Additionally, it is capable of scaling to a computational speed of up to 160 tera-operations per second (TOPS) and a computational density as high as 25.6 TOPS/mm2. Our approach holds significant promise for enabling highly efficient, scalable on-chip optical computing, with broad potential applications in AI and beyond.

physics.optics↗

arXiv2025

Chat3GPP: An Open-Source Retrieval-Augmented Generation Framework for 3GPP Documents

The 3rd Generation Partnership Project (3GPP) documents is key standards in global telecommunications, while posing significant challenges for engineers and researchers in the telecommunications field due to the large volume and complexity of their contents as well as the frequent updates. Large language models (LLMs) have shown promise in natural language processing tasks, but their general-purpose nature limits their effectiveness in specific domains like telecommunications. To address this, we propose Chat3GPP, an open-source retrieval-augmented generation (RAG) framework tailored for 3GPP specifications. By combining chunking strategies, hybrid retrieval and efficient indexing methods, Chat3GPP can efficiently retrieve relevant information and generate accurate responses to user queries without requiring domain-specific fine-tuning, which is both flexible and scalable, offering significant potential for adapting to other technical standards beyond 3GPP. We evaluate Chat3GPP on two telecom-specific datasets and demonstrate its superior performance compared to existing methods, showcasing its potential for downstream tasks like protocol generation and code automation.

cs.CL↗

arXiv2024

HQOD: Harmonious Quantization for Object Detection

Task inharmony problem commonly occurs in modern object detectors, leading to inconsistent qualities between classification and regression tasks. The predicted boxes with high classification scores but poor localization positions or low classification scores but accurate localization positions will worsen the performance of detectors after Non-Maximum Suppression. Furthermore, when object detectors collaborate with Quantization-Aware Training (QAT), we observe that the task inharmony problem will be further exacerbated, which is considered one of the main causes of the performance degradation of quantized detectors. To tackle this issue, we propose the Harmonious Quantization for Object Detection (HQOD) framework, which consists of two components. Firstly, we propose a task-correlated loss to encourage detectors to focus on improving samples with lower task harmony quality during QAT. Secondly, a harmonious Intersection over Union (IoU) loss is incorporated to balance the optimization of the regression branch across different IoU levels. The proposed HQOD can be easily integrated into different QAT algorithms and detectors. Remarkably, on the MS COCO dataset, our 4-bit ATSS with ResNet-50 backbone achieves a state-of-the-art mAP of 39.6%, even surpassing the full-precision one.

cs.CV↗

arXiv2024

Soft Multipath Information-Based UWB Tracking in Cluttered Scenarios: Preliminaries and Validations

In this paper, we investigate ultra-wideband (UWB) localization and tracking in cluttered environments. Instead of mitigating the multipath, we exploit the specular reflections to enhance the localizability and improve the positioning accuracy. With the assistance of the multipath, it is also possible to achieve localization purposes using fewer anchors or when the line-of-sight propagations are blocked. Rather than using single-value distance, angle, or Doppler estimates for the localization, we model the likelihoods of both the line-of-sight and specular multipath components, namely soft multipath information, and propose the multipath-assisted probabilistic UWB tracking algorithm. Experimental results in a cluttered industrial scenario show that the proposed algorithm achieves 46.4 cm and 33.1 cm 90th percentile errors in the cases of 3 and 4 anchors, respectively, which outperforms conventional methods with more than 61.8% improvement given fewer anchors and strong multipath effect.

eess.SP↗

arXiv2024

DiffMap: Enhancing Map Segmentation with Map Prior Using Diffusion Model

Constructing high-definition (HD) maps is a crucial requirement for enabling autonomous driving. In recent years, several map segmentation algorithms have been developed to address this need, leveraging advancements in Bird's-Eye View (BEV) perception. However, existing models still encounter challenges in producing realistic and consistent semantic map layouts. One prominent issue is the limited utilization of structured priors inherent in map segmentation masks. In light of this, we propose DiffMap, a novel approach specifically designed to model the structured priors of map segmentation masks using latent diffusion model. By incorporating this technique, the performance of existing semantic segmentation methods can be significantly enhanced and certain structural errors present in the segmentation outputs can be effectively rectified. Notably, the proposed module can be seamlessly integrated into any map segmentation model, thereby augmenting its capability to accurately delineate semantic information. Furthermore, through extensive visualization analysis, our model demonstrates superior proficiency in generating results that more accurately reflect real-world map layouts, further validating its efficacy in improving the quality of the generated maps.

cs.CV↗

arXiv2024

Only the Curve Shape Matters: Training Foundation Models for Zero-Shot Multivariate Time Series Forecasting through Next Curve Shape Prediction

We present General Time Transformer (GTT), an encoder-only style foundation model for zero-shot multivariate time series forecasting. GTT is pretrained on a large dataset of 200M high-quality time series samples spanning diverse domains. In our proposed framework, the task of multivariate time series forecasting is formulated as a channel-wise next curve shape prediction problem, where each time series sample is represented as a sequence of non-overlapping curve shapes with a unified numerical magnitude. GTT is trained to predict the next curve shape based on a window of past curve shapes in a channel-wise manner. Experimental results demonstrate that GTT exhibits superior zero-shot multivariate forecasting capabilities on unseen time series datasets, even surpassing state-of-the-art supervised baselines. Additionally, we investigate the impact of varying GTT model parameters and training dataset scales, observing that the scaling law also holds in the context of zero-shot multivariate time series forecasting.

cs.LG↗

arXiv2023

Sharp Poincar\'e--Sobolev Inequalities of Choquet--Lorentz Integrals with Respect to Hausdorff Contents on Bounded John Domains

Let $\Omega$ be a bounded John domain in $\mathbb R^n$ with $n\ge 2$, and let $\mathcal{H}_{\infty }^{\delta}$ denote the Hausdorff content of dimension $\delta\in (0,n]$. In this article, the authors prove the Poincar\'e and the Poincar\'e--Sobolev inequalities, with sharp ranges of indices, on Choquet--Lorentz integrals with respect to $\mathcal{H}_{\infty }^{\delta}$ for all continuously differentiable functions on $\Omega$. These results not only extend the recent Poincar\'e and Poincar\'e--Sobolev inequalities to the Choquet--Lorentz integrals, but also provide some endpoint estimates (weak type) in the critical case. One of the main novelties exists in that, to achieve the goals, the authors develop some new tools associated with Choquet--Lorentz integrals on $\mathcal{H}_{\infty }^{\delta}$, such as the fractional Hardy--Littlewood maximal inequality and the Hedberg-type pointwise estimate on the Riesz potential. As an application, the authors obtain the sharp boundedness of the Riesz potential on Choquet--Lorentz integrals. Moreover, even for classical Lorentz integrals, these Poincar\'e and Poincar\'e--Sobolev inequalities are also new.

math.FA↗