arXiv subjects

Guanghui He

Publications and source records attributed to Guanghui He.

At least 19 recordsLinked to original sources

arXiv2026

Algorithm-Architecture Co-Design for Efficient VLA Inference via Speculative Inference and Verification

Vision-Language-Action (VLA) models have demonstrated remarkable capabilities in the field of embodied AI, but their high computational cost and limited predicted action length hinder real-time deployment. Although Dadu-Corki, a dedicated accelerator for efficient embodied AI, has been introduced, it does not exploit the inherent interaction patterns between the robot and its environment, which results in a relatively short predicted action length. We observe that robotic environments naturally alternate between active states-where precise actions are crucial-and inactive states-where actions have limited impact on task success. This insight enables a new scheduling opportunity: long-action-length speculative prediction in inactive states, paired with selective verification in active states. We propose SpecVLA, an algorithm-system co-design framework that adaptively balances action length, inference latency, and task reliability. On the algorithm side, SpecVLA introduces a state-aware VLA inference execution paradigm and a hardware-friendly construction of a smaller verification model (sVLA) using differential residuals and block-wise mixed-precision quantization. On the system side, we develop a heterogeneous architecture consisting of a GPU and a robotic-specific hardware module, along with a speculative dataflow that decouples VLA and sVLA through parallel execution. Comprehensive evaluations on OpenVLA and RDT across LIBERO and ManiSkill benchmarks show that SpecVLA reduces end-to-end latency significantly while preserving task success rate. By enabling long-action-length speculative prediction with timely verification, SpecVLA achieves real-time robotic manipulation with both high efficiency and reliability.

cs.RO↗

arXiv2026

RhinoVLA Technical Report

Vision-Language-Action (VLA) models have shown strong potential for robotic manipulation, but real-time deployment on edge hardware remains challenging. In this work, we identify VLM visual and context tokens as a major source of deployment latency: for GEMM-dominated projection operators, computation grows linearly with the number of input tokens when model dimensions are fixed. Motivated by this observation, we propose RhinoVLA, a deployment-oriented VLA model co-designed with the Huixi R1 edge SoC. RhinoVLA adopts a token-efficient Qwen3-VL backbone and a continuous Action Expert, reducing the VLM-side token and computation burden while preserving pretrained multimodal capability. To support cross-robot learning, RhinoVLA further introduces a unified interface that combines View Registry, 72D physical state-action slot space, and robotinstance LoRA, allowing heterogeneous robot observations and action schemas to be aligned under a shared policy. On the deployment side, RhinoVLA is optimized through hardware-aware compilation, mixed-precision execution, and parallel visual encoding. Experiments show that RhinoVLA achieves downstream performance comparable to {\pi}0.5 at a similar parameter scale, while reaching 11.69 Hz end-to-end inference on Huixi R1, meeting the 10 Hz real-time closedloop control target. The project will be open-sourced at https://github.com/HuixiAI/RhinoVLA.

cs.RO↗

arXiv2026

Accelerating MoE with Dynamic In-Switch Computing on Multi-GPUs

Mixture-of-Experts (MoE) has been adopted by many leading large models to reduce computational requirements. However, frequent inter-GPU communication in MoE expert parallelism (EP) becomes a performance challenge. We observe substantial redundant inter-GPU data transfers in MoE that can be potentially addressed by in-switch computing. Unfortunately, the existing solution, NVLink SHARP (NVLS), can only support static collectives with regular patterns, incapable of dynamic communication with irregular patterns in MoE. To bridge the functionality gap, we propose DySHARP, an integral dynamic in-switch computing solution to accelerate MoE, encompassing both communication primitives and communication-aware scheduling: 1) Dynamic multimem addressing co-designs ISA, architecture, and runtime, as a dynamic extension to NVLS, reducing redundant traffic. However, the resulting traffic reduction is inherently asymmetric between two directions, preventing it from directly translating into speedup. 2) Token-centric kernel fusion deeply fuses the dispatch-computation-combine pipeline, resolving this asymmetry to translate traffic reduction into actual speedup. Compared with the state-of-the-art solution, DySHARP achieves up to 1.79$\times$ speedup.

cs.AR↗

arXiv2026

MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems

The Mixture-of-Experts (MoE) architecture is crucial for scaling large language models, but its scalability is severely limited by inter-GPU communication bottlenecks in multi-GPU systems. Although overlapping communication with computation is a widely recognized optimization, its effective deployment still remains challenging, both in terms of performance and programmability. In this work, we identify the root cause as a fundamental abstraction mismatch between MoE's dynamic, irregular token-to-expert mapping and the static, address-centric communication model of modern GPUs, which necessitates a complex software mediation phase to resolve addresses before data transfers, limiting performance and software flexibility. To resolve this, we propose MoE-Hub, a hardware-software co-design that introduces a destination-agnostic communication paradigm. MoE-Hub decouples data transmission from address management, allowing producers to send data immediately after routing using only a logical destination, while address allocation and data-flow orchestration are handled transparently by lightweight hardware in the GPU hub. By hardware-accelerating the entire communication control plane, MoE-Hub enables seamless and transparent overlap. Our evaluation shows that MoE-Hub achieves 1.40x-3.08x per-layer and 1.21x-1.98x end-to-end speedup over state-of-the-art systems.

cs.AR↗

arXiv2026

DS2SC-Agent: A Multi-Agent Automated Pipeline for Rapid Chiplet Model Generation

Constructing behavioral-level chiplet models (e.g., SystemC) is crucial for early-stage heterogeneous architecture exploration. Traditional manual modeling is notoriously time-consuming and error-prone. Recently, Large Language Models (LLMs) have demonstrated immense potential in automating hardware code generation. However, existing LLM-assisted design frameworks predominantly target highly structured or well-defined design specifications. In practical engineering scenarios, raw datasheets typically encompass lengthy, complex, and highly unstructured information. Consequently, reliable code generation directly from these raw datasheets suffers from severe challenges, including context vanishing and logical hallucinations.To overcome this critical bottleneck, this paper proposes DS2SC-Agent(Datasheet-to-SystemC-Agent): the first end-to-end, fully automated generation pipeline that translates raw datasheets directly into SystemC chiplet models. This system establishes a highly efficient multi-agent collaborative framework. By decoupling the intricate modeling tasks, the proposed pipeline orchestrates a fully automated workflow encompassing unstructured long-document parsing, SystemC core code construction, testbench stimulus generation, and adaptive closed-loop debugging. We comprehensively evaluate the proposed framework on representative single-function chiplets across the analog, digital, and radio frequency (RF) domains--specifically, a Limiting Amplifier (LA), a Fast Fourier Transform (FFT) module, and a Power Amplifier (PA). The evaluation demonstrates that our pipeline seamlessly processes complex real-world datasheets to consistently generate functionally correct SystemC models. This provides a highly efficient and reliable paradigm for agile model library construction while drastically minimizing manual intervention.

cs.AR↗

arXiv2026

Spin Relaxometry with Solid-State Defects: Theory, Platforms, and Applications

Spin relaxometry using solid-state spin defects, such as the diamond nitrogen-vacancy (NV) center, probes dynamical processes by measuring how environmental fluctuations enhance the spin relaxation rate. In the weak-coupling limit, relaxation rates sample the transverse magnetic-noise power spectral density through a sensor-specific filter function, turning the defect into a local, frequency-selective noise spectrometer. This review bridges theory and experiment, clarifying how measured relaxation rates map onto noise spectra and how near-field geometry shapes the response. We highlight representative applications across condensed-matter physics, chemical and biological sensing, and relaxometry-based magnetic-resonance spectroscopy. We conclude with emerging opportunities and key challenges.

cond-mat.mes-hall↗

arXiv2025

SC-LDPC Codes Over $\mathbb{F}_q$: Minimum Distance, Decoding Analysis and Threshold Saturation

We investigate random spatially coupled low-density parity-check (SC-LDPC) code ensembles over finite fields. Under different variable-node edge-spreading rules, the random Tanner graphs of several coupled ensembles are defined by multiple independent, uniformly random monomial maps. The two main coupled ensembles considered are referred to as the standard coupled ensemble and the improved coupled ensemble. We prove that both coupled ensembles exhibit asymptotically good minimum distance and minimum stopping set size. Theoretical and numerical results show that the improved coupled ensemble can achieve better distance performance than the standard coupled ensemble. We introduce the essential preliminaries and analytical tools needed to analyze the iterative decoding threshold of coupled ensembles over any finite field. We consider a class of memoryless channels with special symmetry, termed q-ary input memoryless symmetric channels (QMSCs), and show that, for these channels, the distribution of channel messages (in form of probability vectors) likewise exhibits this symmetry. Consequently, we define symmetric probability measures and their reference measures on a finite-dimensional probability simplex, analyze their foundational properties and those of their linear functionals, endow their respective spaces with metric topologies, and conduct an in-depth study of their degradation theory. Based on our analytical framework, we establish a universal threshold saturation result for both of the coupled ensembles over a q-ary finite field on QMSCs. Specifically, as the coupling parameters increase, the belief-propagation threshold of a coupled system saturates to a well-defined threshold that depends only on the underlying ensemble and the channel family.

cs.IT↗

arXiv2025

SD-Acc: Accelerating Stable Diffusion through Phase-aware Sampling and Hardware Co-Optimizations

The emergence of diffusion models has significantly advanced generative AI, improving the quality, realism, and creativity of image and video generation. Among them, Stable Diffusion (StableDiff) stands out as a key model for text-to-image generation and a foundation for next-generation multi-modal algorithms. However, its high computational and memory demands hinder inference speed and energy efficiency. To address these challenges, we identify three core issues: (1) intensive and often redundant computations, (2) heterogeneous operations involving convolutions and attention mechanisms, and (3) diverse weight and activation sizes. We present SD-Acc, a novel algorithm and hardware co-optimization framework. At the algorithm level, we observe that high-level features in certain denoising phases show significant similarity, enabling approximate computation. Leveraging this, we propose an adaptive, phase-aware sampling strategy that reduces compute and memory loads. This framework automatically balances image quality and complexity based on the StableDiff model and user requirements. At the hardware level, we design an address-centric dataflow to efficiently handle heterogeneous operations within a simple systolic array. We address the bottleneck of nonlinear functions via a two-stage streaming architecture and a reconfigurable vector processing unit. Additionally, we implement adaptive dataflow optimizations by combining dynamic reuse and operator fusion tailored to StableDiff workloads, significantly reducing memory access. Across multiple StableDiff models, our method achieves up to a 3x reduction in computational demand without compromising image quality. Combined with our optimized hardware accelerator, SD-Acc delivers higher speed and energy efficiency than traditional CPU and GPU implementations.

cs.AR↗

arXiv2025

VEDA: Efficient LLM Generation Through Voting-based KV Cache Eviction and Dataflow-flexible Accelerator

Large Language Models (LLMs) excel in natural language processing tasks but pose significant computational and memory challenges for edge deployment due to their intensive resource demands. This work addresses the efficiency of LLM inference by algorithm-hardware-dataflow tri-optimizations. We propose a novel voting-based KV cache eviction algorithm, balancing hardware efficiency and algorithm accuracy by adaptively identifying unimportant kv vectors. From a dataflow perspective, we introduce a flexible-product dataflow and a runtime reconfigurable PE array for matrix-vector multiplication. The proposed approach effectively handles the diverse dimensional requirements and solves the challenges of incrementally varying sequence lengths. Additionally, an element-serial scheduling scheme is proposed for nonlinear operations, such as softmax and layer normalization (layernorm). Results demonstrate a substantial reduction in latency, accompanied by a significant decrease in hardware complexity, from O(N) to O(1). The proposed solution is realized in a custom-designed accelerator, VEDA, which outperforms existing hardware platforms. This research represents a significant advancement in LLM inference on resource-constrained edge devices, facilitating real-time processing, enhancing data privacy, and enabling model customization.

cs.AR↗

arXiv2025

Efficient 3D Gaussian Splatting with Axis-Shared Rasterization and Order-independent Transmittance

3D Gaussian Splatting (3DGS) has emerged as a powerful technique for novel view synthesis, combining high-quality reconstruction with efficient rendering. It has been widely adopted in domains such as AR/VR, robotics, and autonomous driving. However, achieving real-time performance on resource-constrained platforms remains challenging due to strict power and area budgets. Prior accelerators improve hardware performance but still overlook key inefficiencies, including insufficient rasterization efficiency, poor sorting scalability, and pipeline imbalance. This paper presents an architecture-algorithm co-design to address these challenges. First, we propose axis-shared rasterization, which precomputes and reuses common terms along the X- and Y-axes, reducing multiply-and-accumulate (MAC) operations by up to 38% while preserving high parallelism. Second, we develop a novel order-independent transmittance method that removes the need for explicit sorting by leveraging a lightweight multilayer perceptron (MLP) to directly approximate the transmittance of each Gaussian, enabling efficient alpha blending with negligible quality loss. Third, we design a unified reconfigurable PE array that supports both rasterization and MLP inference, sustaining high utilization without costly sorting hardware. Our experiments demonstrate that our design preserves rendering quality while achieving a 1.33 to 1.88x speedup over state-of-the-art 3DGS accelerators. Our code is open source at https://github.com/WangZhican/ISCA26_3DGS_Acc.

cs.GR↗

arXiv2025

Probing Boron Vacancy Defects in hBN via Single Spin Relaxometry

Spin defects in solids offer promising platforms for quantum sensing and memory due to their long coherence times and optical addressability. Here, we integrate a single nitrogen-vacancy (NV) center in diamond with scanning probe microscopy to discover, read out, and spatially map arbitrary spin-based quantum sensors at the nanoscale. Using the boron vacancy ($\mathrm{V}_\mathrm{B}^-$) center in hexagonal boron nitride$\unicode{x2013}$an emerging two-dimensional spin system$\unicode{x2013}$as a model, we detect its electron spin resonance indirectly via changes in the spin relaxation time ($T_1$) of a nearby NV center, eliminating the need for optical excitation or fluorescence detection of the $\mathrm{V}_\mathrm{B}^-$. Cross-relaxation between NV and $\mathrm{V}_\mathrm{B}^-$ ensembles significantly reduces NV $T_1$, enabling quantitative nanoscale mapping of defect densities beyond the optical diffraction limit and clear resolution of hyperfine splitting in isotopically enriched h$^{10}$B$^{15}$N. Our method demonstrates interactions between 3D and 2D spin sensors, establishing NV centers as versatile probes for characterizing otherwise inaccessible spin defects.

cond-mat.mtrl-sci↗

arXiv2025

Quantum noise spectroscopy of superconducting dynamics in thin film Bi$_2$Sr$_2$CaCu$_2$O$_{8+\delta}$

Characterizing the low-energy dynamics of quantum materials is crucial to our understanding of strongly correlated electronic states. Yet, it remains experimentally challenging to investigate such dynamics with high spectroscopic resolution in both frequency and momentum space, particularly in two-dimensional correlated systems. Here, we leverage Nitrogen-Vacancy (NV) centers in diamond as a powerful and non-invasive tool to study thin-film Bi$_2$Sr$_2$CaCu$_2$O$_{8+\delta}$ (BSCCO), revealing several distinct dynamical phenomena across the superconducting phase diagram. At zero magnetic field and low temperatures, NV depolarization ($T_1$) noise spectroscopy captures the low-frequency (GHz-scale) magnetic noise generated by nodal superconducting quasiparticle excitations, in agreement with Bardeen-Cooper-Schrieffer (BCS) mean-field theory. Near the critical temperature $T_c \approx 90$ K, supercurrent-fluctuation-induced noise leads to a sharp reduction of the NV $T_1$. By carefully analyzing the temperature-scaling of $T_1$, we observe clear deviations from the BCS prediction, reflecting the importance of order parameter fluctuations and enabling the determination of both static and dynamical critical exponents. When a small field is applied, we detect a broad and asymmetric reduction of NV $T_1$ near $T_c$; the field-induced smearing of the transition unveils the presence of a vortex liquid phase. Finally, NV decoherence ($T_2$) noise spectroscopy allows us to characterize magnetic noise at even lower MHz-scale frequencies and obtain evidence for complex vortex-solid fluctuations well below $T_c$. Our results establish quantum noise spectroscopy as a versatile platform for probing dynamical phenomena in superconductors, with frequency and length scales complementary to existing techniques.

cond-mat.supr-con↗

arXiv2025

Probing Stress and Magnetism at High Pressures with Two-Dimensional Quantum Sensors

Pressure serves as a fundamental tuning parameter capable of drastically modifying all properties of matter. The advent of diamond anvil cells (DACs) has enabled a compact and tabletop platform for generating extreme pressure conditions in laboratory settings. However, the limited spatial dimensions and ultrahigh pressures within these environments present significant challenges for conventional spectroscopy techniques. In this work, we integrate optical spin defects within a thin layer of two-dimensional (2D) materials directly into the high-pressure chamber, enabling an in situ quantum sensing platform for mapping local stress and magnetic environments up to 4~GPa. Compared to nitrogen-vacancy (NV) centers embedded in diamond anvils, our 2D sensors exhibit around three times stronger response to local stress and provide nanoscale proximity to the target sample in heterogeneous devices. We showcase the versatility of our approach by imaging both stress gradients within the high-pressure chamber and a pressure-driven magnetic phase transition in a room-temperature self-intercalated van der Waals ferromagnet, Cr$_{1+\delta}$Te$_2$. Our work demonstrates an integrated quantum sensing device for high-pressure experiments, offering potential applications in probing pressure-induced phenomena such as superconductivity, magnetism, and mechanical deformation.

cond-mat.mes-hall↗

arXiv2024

Temperature dependent spin-phonon coupling of boron-vacancy centers in hexagonal boron nitride

The negatively charged boron-vacancy center ($\mathrm{V}_{\mathrm{B}}^-$) in hexagonal boron nitride (hBN) has recently emerged as a highly promising quantum sensor. Compared to the nitrogen-vacancy (NV) center in diamond, the change with temperature of the spin transition energy of $\mathrm{V}_{\mathrm{B}}^-$ is more than an order of magnitude larger, making it a potential nanoscale thermometer with superior sensitivity. However, the underlying mechanism of the observed large temperature dependence remains an open question. In this work, using isotopically purified $\mathrm{h}{}^{10}\mathrm{B}{}^{15}\mathrm{N}$, we systematically characterize the zero-field splitting, hyperfine interaction, and spin relaxation time of $\mathrm{V}_{\mathrm{B}}^-$ from 10 to 350$~$K. We carry out first-principle calculations of the $\mathrm{V}_{\mathrm{B}}^-$ spin-phonon interaction and show that a second-order effect from finite-temperature phonon excitations is responsible for the observed changes in experiments. By fitting our experimental results to a physically motivated model, we extract the dominant phonon mode which agrees well with our simulations. Finally, we investigate the dynamic nuclear spin polarization process at cryogenic temperatures. Our results provide key insights in $\mathrm{V}_{\mathrm{B}}^-$ centers and their utilization as nanoscale thermometers and phonon sensors.

quant-ph↗

arXiv2024

Experimental Realization of Discrete Time Quasi-Crystals

Floquet (periodically driven) systems can give rise to unique non-equilibrium phases of matter without equilibrium analogs. The most prominent example is the realization of discrete time crystals. An intriguing question emerges: what other novel phases can manifest when the constraint of time periodicity is relaxed? In this study, we explore quantum systems subjected to a quasi-periodic drive. Leveraging a strongly interacting spin ensemble in diamond, we identify the emergence of long-lived discrete time quasi-crystals. Unlike conventional time crystals, time quasi-crystals exhibit robust sub-harmonic responses at multiple incommensurate frequencies. Furthermore, we show that the multi-frequency nature of the quasi-periodic drive allows for the formation of diverse patterns associated with different discrete time quasi-crystalline phases. Our findings demonstrate the existence of non-equilibrium phases in quasi-Floquet settings, significantly broadening the catalog of novel phenomena in driven many-body quantum systems.

quant-ph↗

arXiv2024

DEFA: Efficient Deformable Attention Acceleration via Pruning-Assisted Grid-Sampling and Multi-Scale Parallel Processing

Multi-scale deformable attention (MSDeformAttn) has emerged as a key mechanism in various vision tasks, demonstrating explicit superiority attributed to multi-scale grid-sampling. However, this newly introduced operator incurs irregular data access and enormous memory requirement, leading to severe PE underutilization. Meanwhile, existing approaches for attention acceleration cannot be directly applied to MSDeformAttn due to lack of support for this distinct procedure. Therefore, we propose a dedicated algorithm-architecture co-design dubbed DEFA, the first-of-its-kind method for MSDeformAttn acceleration. At the algorithm level, DEFA adopts frequency-weighted pruning and probability-aware pruning for feature maps and sampling points respectively, alleviating the memory footprint by over 80%. At the architecture level, it explores the multi-scale parallelism to boost the throughput significantly and further reduces the memory access via fine-grained layer fusion and feature map reusing. Extensively evaluated on representative benchmarks, DEFA achieves 10.1-31.9x speedup and 20.3-37.7x energy efficiency boost compared to powerful GPUs. It also rivals the related accelerators by 2.2-3.7x energy efficiency improvement while providing pioneering support for MSDeformAttn.

cs.AR↗

arXiv2024

Isotope engineering for spin defects in van der Waals materials

Spin defects in van der Waals materials offer a promising platform for advancing quantum technologies. Here, we propose and demonstrate a powerful technique based on isotope engineering of host materials to significantly enhance the coherence properties of embedded spin defects. Focusing on the recently-discovered negatively charged boron vacancy center ($\mathrm{V}_{\mathrm{B}}^-$) in hexagonal boron nitride (hBN), we grow isotopically purified $\mathrm{h}{}^{10}\mathrm{B}{}^{15}\mathrm{N}$ crystals. Compared to $\mathrm{V}_{\mathrm{B}}^-$ in hBN with the natural distribution of isotopes, we observe substantially narrower and less crowded $\mathrm{V}_{\mathrm{B}}^-$ spin transitions as well as extended coherence time $T_2$ and relaxation time $T_1$. For quantum sensing, $\mathrm{V}_{\mathrm{B}}^-$ centers in our $\mathrm{h}{}^{10}\mathrm{B}{}^{15}\mathrm{N}$ samples exhibit a factor of $4$ ($2$) enhancement in DC (AC) magnetic field sensitivity. For additional quantum resources, the individual addressability of the $\mathrm{V}_{\mathrm{B}}^-$ hyperfine levels enables the dynamical polarization and coherent control of the three nearest-neighbor ${}^{15}\mathrm{N}$ nuclear spins. Our results demonstrate the power of isotope engineering for enhancing the properties of quantum spin defects in hBN, and can be readily extended to improving spin qubits in a broad family of van der Waals materials.

quant-ph↗

arXiv2023

Deep Learning-Based Modeling of 5G Core Control Plane for 5G Network Digital Twin

Digital twin serves as a crucial facilitator in the advancement and implementation of emerging technologies within 5G and beyond networks. However, the intricate structure and diverse functionalities of the existing 5G core network, especially the control plane, present challenges in constructing core network digital twins. In this paper, we propose two novel data-driven architectures for modeling the 5G control plane and implement corresponding deep learning models, namely 5GC-Seq2Seq and 5GC-former, based on the Vanilla Seq2Seq model and Transformer decoder respectively. We also present a solution enabling the interconversion of signaling messages and length-limited vectors to construct a dataset. The experiments are based on 5G core network signaling messages collected by the Spirent C50 network tester, encompassing various procedures such as registration, handover, and PDU sessions. The results show that 5GC-Seq2Seq achieves a 99.997\% F1-score (a metric measuring the accuracy of positive samples) in single UE scenarios with a simple structure, but exhibits significantly reduced performance in handling concurrency. In contrast, 5GC-former surpasses 99.999\% F1-score while maintaining robust performance under concurrent UE scenarios by constructing a more complex and highly parallel model. These findings validate that our method accurately replicates the principal functionalities of the 5G core network control plane.

eess.SY↗