arXiv subjects

Junhua Wang

Publications and source records attributed to Junhua Wang.

At least 19 recordsLinked to original sources

arXiv2026

Amplitude-Only FFN Intervention for Tool-Structured LLM Inference Method: Gated Evaluation Protocol, and Cross-Model Empirical Results

Large language models increasingly operate as tool-using agents, where small format, argument, or function-call errors can invalidate otherwise plausible responses. We study inference-time feed-forward network (FFN) intervention as a way to improve structured outputs without retraining model weights. An earlier project-specific approach, Orthogonal Residual Projection (ORP), exposed sensitive SwiGLU FFN sites and non-monotonic energy effects, but its direction-changing operation produced more regressions than repairs in a key diagnostic. We therefore propose Amplitude Gating (AG), which preserves pretrained FFN weight directions and modulates activation magnitudes during decoding. AG separates candidate generation, ranking, and a prospective acceptance/fallback decision. We also introduce Per-Sample Fix-Harm Evaluation (PFHE), a paired reporting protocol that complements native task metrics with fixes, harms, preserved-correct cases, and preserved-wrong cases. On the only cross-position union that passes source-alignment audit, an exploratory offline mixed selector raises the descriptive heterogeneous-scorer Qwen3.5-9B tool-route micro-average from 38.66% to 42.92% (+4.27 percentage points); two Hermes function-call endpoints improve by +7.64 and +7.62 points. The same-output PFHE-format view records 48 fixes, 26 harms, 294 preserved-correct cases, and 2,188 preserved-wrong cases over 2,556 units, with positive paired bootstrap intervals for native and strict effects. Protocol-separated Qwen3-8B and Qwen2.5-7B analyses retain oracle headroom but no positive train-selected fixed tool route. A grouped five-fold RF diagnostic suggests weak nonlinear ranking signal but forces intervention, lacks baseline fallback and paired uncertainty, and is not deployment evidence. The results support model- and task-specific selection with strict fallback, not a universal AG switch.

cs.CL↗

arXiv2026

FedCoE: Bridging Generalization and Personalization via Federated Coordinated Dual-level MoEs

Federated Learning (FL) has emerged as a promising paradigm for privacy-preserving distributed learning. However, existing FL methods face a fundamental challenge. Traditional averaging-based approaches suffer from parameter divergence under non-IID conditions, while personalized FL methods overfit to local data and fail to generalize to new clients (cold-start problem). Mixture-of-Experts naturally addresses this by routing heterogeneous data to specialized experts rather than forcing uniform aggregation. In this paper, we propose FedCoE, a Federated Coordinated dual-level mixture-of-Experts framework that effectively balances global generalization with local personalization. FedCoE maintains multiple independent global expert models on the server and employs a shared gating network to dynamically model client-expert correlations during aggregation, effectively mitigating expert drift and gating inconsistency. To address the cold-start challenge, we introduce an adaptive mechanism that enables new clients to immediately leverage the global expert pool without extensive local training. Extensive experiments demonstrate that FedCoE achieves 78.00% global accuracy and 89.32% personalized accuracy on average, outperforming the baseline by 8.82% and 29.19%, respectively. In cold-start scenarios, FedCoE delivers 77.27% accuracy without any local fine-tuning, outperforming baselines by over 12.54%.

cs.LG↗

arXiv2026

AutoMCU: Feasibility-First MCU Neural Network Customization via LLM-based Multi-Agent Systems

Deploying neural networks on microcontroller units (MCUs) is critical for edge intelligence but remains challenging due to tight memory, storage, and computation constraints. Existing approaches, such as model compression and hardware-aware neural architecture search (HW-NAS), often depend on proxy metrics, incur high search cost, and do not fully bridge the gap between architecture design and verified deployment. This paper presents AutoMCU, a feasibility-first large language model (LLM)-based multi-agent system for automated neural network customization under MCU constraints. Given natural-language task requirements and hardware specifications, AutoMCU iteratively generates structured architecture candidates, filters infeasible designs through vendor toolchain feedback before training, evaluates feasible models under a controlled protocol, and verifies deployability through backend-grounded deployment analysis. AutoMCU includes two key mechanisms: 1) hardware-in-the-loop architecture generation for early elimination of undeployable candidates under RAM and Flash constraints, and 2) state-isolated multi-agent scheduling for stable coordination of proposal, training, evaluation, and deployment stages. Experiments on CIFAR-10 and CIFAR-100 under strict MCU constraints show that AutoMCU achieves competitive accuracy while reducing customization time to about 1--2 hours, compared with hundreds of GPU hours for representative MCU-oriented HW-NAS baselines. Comparisons with ColabNAS and the LLM-based NAS method GENIUS on NAS-Bench-201 further demonstrate the effectiveness and stability of AutoMCU. Real-device deployments on multiple STM32 microcontrollers validate its practical applicability to MCU-scale edge intelligence.

cs.LG↗

arXiv2026

Accelerating Compound LLM Training Workloads with Maestro

Compound LLM training workloads-such as knowledge distillation and multimodal LLM (MLLM) training-are gaining prominence. These typically comprise heterogeneous components differing in parameter scale, execution mode (forward-only or full forward-backward), and sequence length. Besides, component activation can be data-dependent: in MLLM training, modality-specific parts activate only when inputs contain corresponding modalities, causing dynamic computational paths and irregular runtime workloads. Conventional frameworks, designed for monolithic models, cannot handle the dual heterogeneity-static (across components) and dynamic (runtime). By enforcing one-size-fits-all training configurations across components and ignoring input-induced variations, they suffer suboptimal throughput and poor GPU utilization. In this paper, we introduce Maestro, a section-centric training framework that addresses both challenges. Maestro first restructures the workload into a coarse-grained section graph. Each section independently configures its parallelism strategy, micro-batch size, and data-parallel degree-enabling fine-grained, component-aware resource allocation to tackle static heterogeneity. To tackle runtime irregularity, Maestro introduces a wavefront scheduling algorithm that dynamically reorders input samples to orchestrate concurrent section execution while preserving cross-section data dependencies. This maximizes inter-section parallelism and minimizes stalls, boosting hardware utilization. Deployed in production for millions of GPU hours, Maestro reduces GPU consumption by ~40% on key workloads-including knowledge distillation and MLLM training-validating its real-world impact.

cs.DC↗

arXiv2026

Characterizing charge-parity detection based on an offset-charge-tunable transmon qubit via randomized benchmarking

Superconducting qubits are compelling platforms for charge-parity detection and, due to their theoretical sensitivity on the meV energy scale, hold promise for rare event searches. In this work, we realize high-fidelity mapping of charge-parity states onto qubit states using an offset-charge-tunable transmon qubit and efficiently characterize the fidelity of the charge-parity detection via randomized benchmarking. Specifically, a gate control line is applied to control offset charge, allowing us to achieve the single-qubit gate fidelity up to 99.96%. We combine a net-zero-based pulse on the gate line with a spin-echo-based sequence to realize charge-parity mapping, achieving a fidelity of 99.37%. Then, we demonstrate continuous monitoring of the charge-parity state with over 93.4% fidelity at a 4-μs sampling interval. Finally, an error analysis of charge-parity detection is performed, and it is found that qubit readout is currently the largest source of error. We believe this work lays the foundation for future exploration of ultra-low energy particles.

quant-ph↗

arXiv2026

FuxiShuffle: An Adaptive and Resilient Shuffle Service for Distributed Data Processing on Alibaba Cloud

Shuffle exchanges intermediate results between upstream and downstream operators in distributed data processing and is usually the bottleneck due to factors such as small random I/Os and network contention. Several systems have been designed to improve shuffle efficiency, but from our experiences of running ultra-large clusters at Alibaba Cloud MaxCompute platform, we observe that they can not adapt to highly dynamic job characteristics and cluster resource conditions, and their fault tolerance mechanisms are passive and inefficient when failures are inevitable. To tackle their limitations, we design and implement FuxiShuffle as a general data shuffle service for the ultra-large production environment of MaxCompute, featuring good adaptability and efficient failure resilience. Specifically, to achieve good adaptability, FuxiShuffle dynamically selects the shuffle mode based on runtime information, conducts progress-aware scheduling for the downstream workers, and automatically determines the most suitable backup strategy for each shuffle data chunk. To make failure resilience efficient, FuxiShuffle actively ensures data availability with multi-replica failover, prevents memory overflow with careful memory management, and employs an incremental recovery mechanism that does not lose computation progress. Our experiments show that, compared to baseline systems, FuxiShuffle significantly reduces not only end-to-end job completion time but also aggregate resource consumption. Micro experiments suggest that our designs are effective in improving adaptability and failure resilience.

cs.DC↗

arXiv2026

A Pathway to Sub-meV Detection of the Dark Universe: Robust Electron Avalanche in the PN junction at 10 mK

The search for light dark matter and cosmic primordial neutrinos necessitates detectors with sub-millielectronvolt (sub-meV) energy thresholds. While superconducting quantum sensors have approached this sensitivity, they often face significant challenges regarding readout complexity and scalability. To address these limitations, we propose a hybrid Superconductor-Insulator-P-N (S-I-P-N) detector architecture. This concept combines the high sensitivity of superconducting Cooper pair breaking with the massive intrinsic gain of semiconductor electron avalanches. A critical prerequisite for this scheme is operation at millikelvin (mK) temperatures, raising the critical fundamental question of whether silicon PN junctions can sustain avalanche multiplication in a regime where carrier freeze-out is severe. Here, we experimentally validate the critical semiconductor amplification stage of the proposed detector. We demonstrate that Silicon Photomultipliers (SiPMs) retain robust Geiger-mode avalanche capabilities at 10 mK. We report a single-photoelectron gain of order 10$^6$ and a dark count rate as low as 5~mHz/mm$^2$, 7 orders of magnitude lower than at room temperature. These results confirm the viability of high-gain semiconductor readout in the deep cryogenic regime, clearing the primary obstacle regarding the semiconductor component for the realization of scalable, sub-meV threshold S-I-P-N detectors.

physics.ins-det↗

arXiv2025

Novel Light Dark Matter Detection with Quantum Parity Detector Using Qubit Arrays

We present the design and the sensitivity reach of the Qubit-based Light Dark Matter detection experiment. We propose the novel two-chip design to reduce signal dissipation, with quantum parity measurement to enhance single-phonon detection sensitivity. We demonstrate the performance of the detector with full phonon and quasiparticle simulations. The experiment is projected to detect $\gtrsim 30$ meV energy deposition with nearly $100\%$ efficiency and high energy resolution. The sensitivity to $m_χ\gtrsim 0.01$ MeV dark matter scattering cross section is expected to be advanced by orders of magnitude for both light and heavy mediators, and similar improvements will be achieved for axion and dark photon absorption in the $0.04$-$0.2$ eV mass range.

hep-ph↗

arXiv2025

Cosmic-ray-induced correlated errors in superconducting qubit array

Correlated errors may devastate quantum error corrections that are necessary for the realization of fault-tolerant quantum computation. Recent experiments with superconducting qubits indicate that they can arise from quasiparticle (QP) bursts induced by cosmic-ray muons and γ-rays. Here, we use charge-parity jump and bit flip for monitoring QP bursts and two muon detectors in the dilution refrigerator for detecting muon events. We directly observe QP bursts leading to correlated errors that are induced solely by muons and separate the contributions of muons and γ-rays. We further investigate the dynamical process of QP burst and the impact of QP trapping on correlated errors and particle detection. The proposed method, which monitors multiqubit simultaneous charge-parity jumps, has high sensitivity to QP burst and may find applications for the detection of cosmic-ray particles, low-mass dark matter, and far-infrared photons.

quant-ph↗

arXiv2024

Mapping Topology-Localization Phase Diagram with Quasiperiodic Disorder Using a Programmable Superconducting Simulator

We explore topology-localization phase diagram by simulating one-dimensional Su-Schrieffer-Heeger (SSH) model with quasiperiodic disorder using a programmable superconducting simulator. We experimentally map out and identify various trivial and topological phases with extended, critical, and localized bulk states. We find that with increasing disorder strength, some extended states can be first replaced by localized states and then by critical states before the system finally becomes fully localized. The critical states exhibit typical features such as multifractality and self-similarity, which lead to surprisingly rich phases with different types of mobility edges and scaling behaviors on the phase boundaries. Our results shed new light on the investigation of the topological and localization phenomena in condensed-matter physics.

quant-ph↗

arXiv2021

Green Internet of Vehicles (IoV) in the 6G Era: Toward Sustainable Vehicular Communications and Networking

As one of the most promising applications in future Internet of Things, Internet of Vehicles (IoV) has been acknowledged as a fundamental technology for developing the Intelligent Transportation Systems in smart cities. With the emergence of the sixth generation (6G) communications technologies, massive network infrastructures will be densely deployed and the number of network nodes will increase exponentially, leading to extremely high energy consumption. There has been an upsurge of interest to develop the green IoV towards sustainable vehicular communication and networking in the 6G era. In this paper, we present the main considerations for green IoV from five different scenarios, including the communication, computation, traffic, Electric Vehicles (EVs), and energy harvesting management. The literatures relevant to each of the scenarios are compared from the perspective of energy optimization (e.g., with respect to resource allocation, workload scheduling, routing design, traffic control, charging management, energy harvesting and sharing, etc.) and the related factors affecting energy efficiency (e.g., resource limitation, channel state, network topology, traffic condition, etc.). In addition, we introduce the potential challenges and the emerging technologies in 6G for developing green IoV systems. Finally, we discuss the research trends in designing energy-efficient IoV systems.

cs.NI↗

arXiv2021

Vacuum-gap transmon qubits realized using flip-chip technology

Significant progress has been made in building large-scale superconducting quantum processors based on flip-chip technology. In this work, we use the flip-chip technology to realize a modified transmon qubit, donated as the "flipmon", whose large shunt capacitor is replaced by a vacuum-gap parallel plate capacitor. To further reduce the qubit footprint, we place one of the qubit pads and a single Josephson junction on the bottom chip and the other pad on the top chip which is galvanically connected with the single Josephson junction through an indium bump. The electric field participation ratio can arrive at nearly 53% in air when the vacuum-gap is about 5 microns, and thus potentially leading to a lower dielectric loss. The coherence times of the flipmons are measured in the range of 30-60 microseconds, which are comparable with that of traditional transmons with similar fabrication processes. The electric field simulation indicates that the metal-air interface's participation ratio increases significantly and may dominate the qubit's decoherence. This suggests that more careful surface treatment needs to be considered. No evidence shows that the indium bumps inside the flipmons cause significant decoherence. With well-designed geometry and good surface treatment, the coherence of the flipmons can be further improved.

quant-ph↗

arXiv2021

Transmon qubit with relaxation time exceeding 0.5 milliseconds

By using the dry etching process of tantalum (Ta) film, we had obtained transmon qubit with the best lifetime (T1) 503 us, suggesting that the dry etching process can be adopted in the following multi-qubit fabrication with Ta film. We also compared the relaxation and coherence times of transmons made with different materials (Ta, Nb and Al) with the same design and fabrication processes of Josephson junction, we found that samples prepared with Ta film had the best performance, followed by those with Al film and Nb film. We inferred that the reason for this difference was due to the different loss of oxide materials located at the metal-air interface.

quant-ph↗

arXiv2020

Realisation of adiabatic and diabatic CZ gates in superconducting qubits coupled with a tunable coupler

High fidelity two-qubit gates are fundamental for scaling up the superconducting number. We use two qubits coupled via a frequency-tunable coupler which can adjust the coupling strength, and demonstrate the CZ gate using two different schemes, adiabatic and diabatic methods. The Clifford based Randomized Benchmarking (RB) method is used to assess and optimize the CZ gate fidelity. The fidelity of adiabatic and diabatic CZ gates are 99.53(8)% and 98.72(2)%, respectively. We also analyze the errors induced by the decoherence. Comparing to 30 ns duration time of adiabatic CZ gate, the duration time of diabatic CZ gate is 19 ns, revealing lower incoherence error rate $r'_{\rm{incoherent, int}}$ = 0.0197(5) than $r_{\rm{incoherent, int}}$ = 0.0223(3).

quant-ph↗

arXiv2020

Characterize noise correlation and enhance coherence via qubit motion

The identification of spacial noise correlation is of critical importance in developing error-corrected quantum devices, but it has barely been studied so far. In this work, we utilize an effective new method called qubit motion, to efficiently determine the noise correlations between any pair of qubits in a 7-qubit superconducting quantum system. The noise correlations between the same pairs of qubits are also investigated when the qubits are at distinct operating frequencies. What's more, in this multi-qubit system with the presence of noise correlations, we demonstrate the enhancing effect of qubit motion on the coherence of logic qubit, and we propose a Motion-CPMG operation sequence to more efficiently protect the logic state from decoherence, which is experimentally demonstrated to extend the decoherence time of logic qubit by nearly one order of magnitude.

quant-ph↗

arXiv2020

On tunnel numbers of a cable knot and its companion

Let $K$ be a nontrivial knot in $S^{3}$ and $t(K)$ its tunnel number. For any $(p\geq 2,q)$-slope in the torus boundary of a closed regular neighborhood of $ K$ in $S^{3}$, denoted by $K^{\star}$, it is a nontrivial cable knot in $S^{3}$. Though $t(K^{\star})\leq t(K)+1$, Example 1.1 in Section 1 shows that in some case, $ t(K^{\star})\leq t(K)$. So it is interesting to know when $t(K^{\star})= t(K)+1$. After using some combinatorial techniques, we prove that (1) for any nontrivial cable knot $K^{\star}$ and its companion $K$, $t(K^{\star})\geq t(K)$; (2) if either $K$ admits a high distance Heegaard splitting or $p/q$ is far away from a fixed subset in the Farey graph, then $t(K^{\star})= t(K)+1$. Using the second conclusion, we construct a satellite knot and its companion so that the difference between their tunnel numbers is arbitrary large.

math.GT↗

arXiv2019

Evidence of Majorana Zero Modes in Josephson Trijunctions

In search of fault-tolerant topological quantum computation (TQC), zero-bias conductance peak as a necessary signature of Majorana zero modes (MZMs) has been observed in a number of solid-state systems. Here, we present the signature of MZMs from a phase-sensitive experiment on Josephson trijunctions constructed on the surface of three-dimensional topological insulators. We observed that the minigap at the center of the trijunction is protected to close over extended regions in phase space, supporting in principle the Majorana phase diagram proposed by Fu and Kane in 2008. Our study paves the way for further braiding MZMs and exploring TQC on a scalable two-dimensional platform.

cond-mat.mes-hall↗

arXiv2018

Protected gap closing in Josephson junctions constructed on Bi$_2$Te$_3$ surface

On the road of searching for Majorana zero modes (MZMs) in topological insulator-based Josephson junctions, a highly-sought signature is the protected full transparency of electron transport through the junctions due to the existence of the MZMs, associated with complete gap closing between the electron-like and hole-like Andreev bound states (ABSs). Here, we present direct experimental evidence of gap closing and full transparency in single Josephson junctions constructed on the surface of three-dimensional topological insulator (3D TI) Bi$_2$Te$_3$. Our results demonstrate that the 2D surface of 3D TIs provides a promising platform for hosting and manipulating MZMs.

cond-mat.mes-hall↗