arXiv subjects

Zewen Sun

Publications and source records attributed to Zewen Sun.

13 recordsLinked to original sources

arXiv2026

RMGS-SLAM: Real-time Multi-sensor Gaussian Splatting SLAM

Achieving real-time Simultaneous Localization and Mapping (SLAM) based on 3D Gaussian splatting (3DGS) in large-scale real-world environments remains challenging, as existing methods still struggle to jointly achieve low-latency pose estimation, continuous 3D Gaussian reconstruction, and long-term global consistency. In this paper, we present a tightly coupled LiDAR-Inertial-Visual 3DGS-based SLAM framework for real-time pose estimation and photorealistic mapping in large-scale real-world scenes. The system executes state estimation and 3D Gaussian primitive initialization in parallel with global Gaussian optimization, enabling continuous dense mapping. To improve Gaussian initialization quality and accelerate optimization convergence, we introduce a cascaded strategy that combines feed-forward predictions with geometric priors derived from voxel-based principal component analysis. To enhance global consistency, we perform loop closure directly on the optimized global Gaussian map by estimating loop constraints through Gaussian-based Generalized Iterative Closest Point registration, followed by pose-graph optimization. We also collect challenging large-scale looped outdoor sequences with hardware-synchronized LiDAR-camera-IMU and ground-truth trajectories for realistic evaluation. Extensive experiments on both public datasets and our dataset demonstrate that the proposed method achieves a state of the art among real-time efficiency, localization accuracy, and rendering quality across diverse real-world scenes.

cs.RO↗

arXiv2026

Wichmann-Kroll Correction in Muonic Atoms and Hydrogen-Like Electronic Ions: a Comparative Study of Two Methods

Wichmann-Kroll corrections are calculated in both hydrogen-like electronic ions and muonic systems ($Z = \{36$--$92\}$) using two independent methods. The Gaussian finite basis set approach, enhanced with dual basis construction, analytical large-distance corrections, and $B$-spline representations, provides computational efficiency. The Green function method, based on semi-analytical construction from Dirac solutions with Fermi nuclear charge distributions, offers higher systematic accuracy and freedom from basis-dependent artifacts. Results are consistent with the literature values, providing reliable reference data for precision spectroscopy of exotic atoms.

physics.atom-ph↗

arXiv2025

Relativistic recoil as a key to the fine-structure puzzle in muonic $^{90}\text{Zr}$

The long-standing fine-structure anomaly in muonic $^{90}$Zr is resolved through a rigorous treatment of the relativistic-recoil effect. From a fit of ab initio QED calculations of the muonic $^{90}$Zr spectrum to precision measurements performed four decades ago, we extract a significantly more precise root-mean-square (rms) charge radius with 6-fold improvement in quality of the fit. A 2-parameter Fermi (2pF) distribution is assumed to model the nuclear charge density and yields a best-fit value of rms charge radius of $r_\text{rms}[^{90}\text{Zr}]=4.2732(7)$ fm ($\chi^2 /{\text{DoF}} = 0.995$), in agreement with the previous muonic spectroscopy value, but a factor $6$ more precise, and 3$\sigma$ larger than the accepted literature value. Additionally, the same analysis has been performed for $^{120}$Sn, where the extracted value of $r_\text{rms}[^{120}\text{Sn}]=4.6518(34)$ fm ($\chi^2 /{\text{DoF}} = 0.88$) is consistent with the accepted value. These results confirm our assumption that the muonic fine-structure puzzle arose from an incomplete treatment of QED effects and their uncertainties.

physics.atom-ph↗

arXiv2025

Position-Prior-Guided Network for System Matrix Super-Resolution in Magnetic Particle Imaging

Magnetic Particle Imaging (MPI) is a novel medical imaging modality. One of the established methods for MPI reconstruction is based on the System Matrix (SM). However, the calibration of the SM is often time-consuming and requires repeated measurements whenever the system parameters change. Current methodologies utilize deep learning-based super-resolution (SR) techniques to expedite SM calibration; nevertheless, these strategies do not fully exploit physical prior knowledge associated with the SM, such as symmetric positional priors. Consequently, we integrated positional priors into existing frameworks for SM calibration. Underpinned by theoretical justification, we empirically validated the efficacy of incorporating positional priors through experiments involving both 2D and 3D SM SR methods.

cs.CV↗

arXiv2025

GAPS: A Clinically Grounded, Automated Benchmark for Evaluating AI Clinicians

Current benchmarks for AI clinician systems, often based on multiple-choice exams or manual rubrics, fail to capture the depth, robustness, and safety required for real-world clinical practice. To address this, we introduce the GAPS framework, a multidimensional paradigm for evaluating Grounding (cognitive depth), Adequacy (answer completeness), Perturbation (robustness), and Safety. Critically, we developed a fully automated, guideline-anchored pipeline to construct a GAPS-aligned benchmark end-to-end, overcoming the scalability and subjectivity limitations of prior work. Our pipeline assembles an evidence neighborhood, creates dual graph and tree representations, and automatically generates questions across G-levels. Rubrics are synthesized by a DeepResearch agent that mimics GRADE-consistent, PICO-driven evidence review in a ReAct loop. Scoring is performed by an ensemble of large language model (LLM) judges. Validation confirmed our automated questions are high-quality and align with clinician judgment (90% agreement, Cohen's Kappa 0.77). Evaluating state-of-the-art models on the benchmark revealed key failure modes: performance degrades sharply with increased reasoning depth (G-axis), models struggle with answer completeness (A-axis), and they are highly vulnerable to adversarial perturbations (P-axis) as well as certain safety issues (S-axis). This automated, clinically-grounded approach provides a reproducible and scalable method for rigorously evaluating AI clinician systems and guiding their development toward safer, more reliable clinical practice. The benchmark dataset GAPS-NSCLC-preview and evaluation code are publicly available at https://huggingface.co/datasets/AQ-MedAI/GAPS-NSCLC-preview and https://github.com/AQ-MedAI/MedicalAiBenchEval.

cs.CL↗

arXiv2025

DoPI: Doctor-like Proactive Interrogation LLM for Traditional Chinese Medicine

Enhancing interrogation capabilities in Traditional Chinese Medicine (TCM) diagnosis through multi-turn dialogues and knowledge graphs presents a significant challenge for modern AI systems. Current large language models (LLMs), despite their advancements, exhibit notable limitations in medical applications, particularly in conducting effective multi-turn dialogues and proactive questioning. These shortcomings hinder their practical application and effectiveness in simulating real-world diagnostic scenarios. To address these limitations, we propose DoPI, a novel LLM system specifically designed for the TCM domain. The DoPI system introduces a collaborative architecture comprising a guidance model and an expert model. The guidance model conducts multi-turn dialogues with patients and dynamically generates questions based on a knowledge graph to efficiently extract critical symptom information. Simultaneously, the expert model leverages deep TCM expertise to provide final diagnoses and treatment plans. Furthermore, this study constructs a multi-turn doctor-patient dialogue dataset to simulate realistic consultation scenarios and proposes a novel evaluation methodology that does not rely on manually collected real-world consultation data. Experimental results show that the DoPI system achieves an accuracy rate of 84.68 percent in interrogation outcomes, significantly enhancing the model's communication ability during diagnosis while maintaining professional expertise.

cs.AI↗

arXiv2025

SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Acceleration

Vision-Language-Action (VLA) models have attracted increasing attention for their strong control capabilities. However, their high computational cost and low execution frequency hinder their suitability for real-time tasks such as robotic manipulation and autonomous navigation. Existing VLA acceleration methods primarily focus on structural optimization, overlooking the fact that these models operate in sequential decision-making environments. As a result, temporal redundancy in sequential action generation and spatial redundancy in visual input remain unaddressed. To this end, we propose SP-VLA, a unified framework that accelerates VLA models by jointly scheduling models and pruning tokens. Specifically, we design an action-aware model scheduling mechanism that reduces temporal redundancy by dynamically switching between VLA model and a lightweight generator. Inspired by the human motion pattern of focusing on key decision points while relying on intuition for other actions, we categorize VLA actions into deliberative and intuitive, assigning the former to the VLA model and the latter to the lightweight generator, enabling frequency-adaptive execution through collaborative model scheduling. To address spatial redundancy, we further develop a spatio-semantic dual-aware token pruning method. Tokens are classified into spatial and semantic types and pruned based on their dual-aware importance to accelerate VLA inference. These two mechanisms work jointly to guide the VLA in focusing on critical actions and salient visual information, achieving effective acceleration while maintaining high accuracy. Extensive experiments show that our method achieves 1.5$\times$ lossless acceleration in LIBERO and 2.4$\times$ in SimplerEnv, with up to 6% average performance gain. Inference frequency and latency improve by 2.2$\times$ in SimplerEnv and 1.4$\times$ in LIBERO.

cs.CV↗

arXiv2025

$^{208}$Pb nuclear charge radius revisited: closing the fine-structure-anomaly gap

A comprehensive reevaluation of the root-mean-square nuclear charge radius is presented for the doubly magic $^{208}$Pb extracted from muonic spectroscopy measurements. By integrating rigorous theoretical quantum electrodynamics calculations, state-of-the-art numerical methods, and a systematic reanalysis of the uncertainties, we reduced the long-standing muonic fine-structure anomaly and improved the goodness of fit by a factor of twenty. The resulting value of 5.5062(5)~fm for a Fermi distribution is fairly consistent with the previously reported muonic spectroscopy value, and three standard deviations larger than the commonly used compilation data, which indicates that the current value and its uncertainty could be significantly underestimated. Attributing the remaining discrepancy to theory errors which can not be rigorously calculated we suggest the rms charge radius with reduced model dependence to be 5.5062(17) fm. This work sets an improved benchmark for charge radius extraction in heavy nuclei and paves a path for systematic reevaluations across the nuclear chart.

physics.atom-ph↗

arXiv2025

OODD: Test-time Out-of-Distribution Detection with Dynamic Dictionary

Out-of-distribution (OOD) detection remains challenging for deep learning models, particularly when test-time OOD samples differ significantly from training outliers. We propose OODD, a novel test-time OOD detection method that dynamically maintains and updates an OOD dictionary without fine-tuning. Our approach leverages a priority queue-based dictionary that accumulates representative OOD features during testing, combined with an informative inlier sampling strategy for in-distribution (ID) samples. To ensure stable performance during early testing, we propose a dual OOD stabilization mechanism that leverages strategically generated outliers derived from ID data. To our best knowledge, extensive experiments on the OpenOOD benchmark demonstrate that OODD significantly outperforms existing methods, achieving a 26.0% improvement in FPR95 on CIFAR-100 Far OOD detection compared to the state-of-the-art approach. Furthermore, we present an optimized variant of the KNN-based OOD detection framework that achieves a 3x speedup while maintaining detection performance.

cs.CV↗

arXiv2024

Scaling Inter-procedural Dataflow Analysis on the Cloud

Apart from forming the backbone of compiler optimization, static dataflow analysis has been widely applied in a vast variety of applications, such as bug detection, privacy analysis, program comprehension, etc. Despite its importance, performing interprocedural dataflow analysis on large-scale programs is well known to be challenging. In this paper, we propose a novel distributed analysis framework supporting the general interprocedural dataflow analysis. Inspired by large-scale graph processing, we devise dedicated distributed worklist algorithms for both whole-program analysis and incremental analysis. We implement these algorithms and develop a distributed framework called BigDataflow running on a large-scale cluster. The experimental results validate the promising performance of BigDataflow -- BigDataflow can finish analyzing the program of millions lines of code in minutes. Compared with the state-of-the-art, BigDataflow achieves much more analysis efficiency.

cs.PL↗

arXiv2023

Nuclear Deformation Effects in the Spectra of Highly Charged Ions

Nuclear deformation effects are theoretically investigated in terms of deformation corrections of the electronic binding and transition energies, $g$ factor, and hyperfine splitting constant. By solving the Dirac equation twice, with the nuclear potential calculated from Fermi and deformed Fermi nuclear density distributions, we separate the deformation effect in binding energies and wavefunctions. The parameters for both models are determined from experimental data. The considered corrections are of interest for spectral analysis and are numerically calculated for the widest possible range of nuclei, consisting over 1100 different samples. The subtleties between different sources of measured data and the corresponding results are discussed. In addition, the importance of deformation effects for the search of new physics with singly-charged ions is examined.

physics.atom-ph↗

arXiv2023

Investigations of 2D ion crystals in a hybrid optical cavity trap for quantum information processing

We numerically investigate a hybrid trapping architecture for 2D ion crystals using static electrode voltages and optical cavity fields for in-plane and out-of-plane confinements, respectively. By studying the stability of 2D crystals against 2D-3D structural phase transitions, we identify the necessary trapping parameters for ytterbium ions. Multiple equilibrium configurations for 2D crystals are possible, and we analyze their stability by estimating potential barriers between them. We find that scattering to anti-trapping states limits the trapping lifetime, which is consistent with recent experiments employing other optical trapping architectures. These 2D ion crystals offer an excellent platform for quantum simulation of frustrated spin systems, benefiting from their 2D triangular lattice structure and phonon-mediated spin-spin interactions. Quantum information processing with tens of ions is feasible in this scheme with current technologies.

quant-ph↗

arXiv2021

Manipulating phonons of a trapped-ion system using optical tweezers

We propose an experimental architecture where an array of optical tweezers affords site-dependent control over the confining potential of a conventional radio-frequency ion trap. The site-dependent control enables programmable manipulation of phonon modes of ions, with many potential applications in quantum information processing (QIP) and thermodynamics. We describe protocols for programming the array of optical tweezers to attain a set of target phonon modes with high accuracy. We propose applications of such controls in simulating quantum thermodynamics of a particle of programmable effective mass via Jarzynski's equality and improving the efficiency of sympathetic cooling and quantum logic gates in a multi-species ion system of disparate masses. We discuss the required optical parameters in a realistic ion trap system and potential adverse effects of optical tweezers in QIP. Our scheme extends the utility of trapped-ions as a platform for quantum computation and simulation.

quant-ph↗