arXiv subjects

Jun Ding

Publications and source records attributed to Jun Ding.

At least 19 recordsLinked to original sources

arXiv2026

Towards general embodied intelligence: integrating large language models, knowledge bases, and reasoning capabilities to build the next generation of AI agents

The convergence of large language models (LLMs), structured knowledge bases (KBs), and reasoning ability (RA) presents a promising trajectory toward general embodied intelligence (GEI). This paper reviews the evolution of LLM-centered intelligent systems, emphasising their integration with knowledge representation, logical reasoning, and physical embodiment. We analyse LLM architectures, pre-training methods, and inference mechanisms, along with their interaction with external knowledge sources and structured reasoning frameworks. Furthermore, we examine embodied intelligence (EI) paradigms wherein agents learn and act in physical environments. To synthesise these dimensions, we present a conceptual framework that illustrates the synergy among LLMs, KBs, RA, and embodiment, serving as a guiding model for perception, reasoning, and action rather than an implemented engineering architecture. To advance toward GEI, we identify five key challenges: efficient LLM deployment, closed-loop knowledge integration, hybrid symbolic-neural reasoning, perception-action grounding, and continual learning. This survey provides a comprehensive roadmap for developing adaptive, multimodal agents capable of operating in complex, dynamic settings.

cs.AI↗

arXiv2026

Accessing the HQET B-Meson Shape Function from a LaMET Quasi-Shape Function

The shape function and the light-cone distribution amplitude of heavy meson jointly characterize the nonperturbative structure of the heavy meson on the light-cone, with the former being essential for theoretical predictions of inclusive decays and the latter for exclusive decays. While first-principles lattice QCD results for the heavy meson LCDA have become available in recent years, lattice results for the shape function remain absent. In this work, we establish a two-step factorization scheme -- known as the HQLaMET framework -- for computing the $B$-meson shape function on the lattice, which fully disentangles the effects of the disparate scales $P_B^z$, $m_b$, and $\Lambda_{\textrm{QCD}}$. For illustration, starting from a phenomenological model for the shape function in HQET, we provide a graphical presentation of the entire procedure of this framework. The results of the current work lay the foundation for nonperturbative lattice QCD determinations of the shape function in the near future.

hep-ph↗

arXiv2026

Engineering Hybrid Resonances in Nanophotonics

Hybridization of resonances is known to overcome inherent limitations of individual systems, enabling advanced functionalities and applications. Here we discuss hybrid plasmonic-Mie resonators that emerged recently as a promising direction in advancing nanophotonic structures by synergistically combining the strong near-field enhancement of plasmonic components with the low-loss, multipolar resonances of dielectric Mie elements. We review the recent progress in the field, encompassing the fundamental physical principles, structural design strategies, material platforms, computational optimization approaches, and representative device implementations. Our discussion starts by evaluating the complementary characteristics of plasmonic and Mie resonances followed by a description of the coupling between these resonances in order to boost light-matter interactions. Afterward, we explore the performance of efficient hybrid resonators for different application areas. Apart from the conventional metal-dielectric systems, we consider the recent class of epsilon-near-zero (ENZ) materials, which can provide unique advantages in terms of field localization, phase engineering, and energy flow management in the vicinity of zero-permittivity conditions, offering more flexibility in designing hybrid nano-optical devices. Lastly, we point out potential research avenues aiming to improve functional and efficient nanophotonic devices, especially those involving emerging topological material systems, such as Sb2Te3, Bi2Te3, Bi2Se3, combining plasmonic amplification, dielectric confinement, and spin-dependent optical behavior.

physics.optics↗

arXiv2026

Summarize Before You Speak with ARACH: A Training-Free Inference-Time Plug-In for Enhancing LLMs via Global Attention Reallocation

Large language models (LLMs) achieve remarkable performance, yet further gains often require costly training. This has motivated growing interest in post-training techniques-especially training-free approaches that improve models at inference time without updating weights. Most training-free methods treat the model as a black box and improve outputs via input/output-level interventions, such as prompt design and test-time scaling through repeated sampling, reranking/verification, or search. In contrast, they rarely offer a plug-and-play mechanism to intervene in a model's internal computation. We propose ARACH(Attention Reallocation via an Adaptive Context Hub), a training-free inference-time plug-in that augments LLMs with an adaptive context hub to aggregate context and reallocate attention. Extensive experiments across multiple language modeling tasks show consistent improvements with modest inference overhead and no parameter updates. Attention analyses further suggest that ARACH mitigates the attention sink phenomenon. These results indicate that engineering a model's internal computation offers a distinct inference-time strategy, fundamentally different from both prompt-based test-time methods and training-based post-training approaches.

cs.CL↗

arXiv2026

Data-Driven Flow Initialization Framework for CFD Acceleration of Underwater Vehicle in Vertical-Plane Oblique Motion

Accurate prediction of flow fields around underwater vehicles undergoing vertical-plane oblique motions is critical for hydrodynamic analysis, but it often requires computationally expensive CFD simulations. This study proposes a Data-Driven Flow Initialization (DDFI) framework that accelerates CFD simulation by integrating deep neural network (DNN) to predict full-domain flow fields. Using the suboff hull under various inlet velocities and angles of attack as an example, a DNN is trained to predict velocity, pressure, and turbulent quantities based on mesh geometry, operating conditions, and hybrid vectors. The DNN can provide reasonably accurate predictions with a relative error about 3.3%. To enhance numerical accuracy while maintaining physical consistency, the DNN-predicted flow fields are utilized as initial solutions for the CFD solver, achieving up to 3.5-fold and 2.0-fold speedup at residual thresholds of 5*10^(-6)and 5*10^(-8), respectively. This method maintains physical consistency by refining neural network outputs via traditional CFD solvers, balancing computational efficiency and accuracy. Notably, reducing the size of training set does not exert an essential impact on acceleration performance. Besides, this method exhibits cross-mesh generalization capability. In general, this proposed hybrid approach offers a new pathway for high-fidelity and efficient full-domain flow field predictions around complex underwater vehicles.

physics.flu-dyn↗

arXiv2025

GCA-ResUNet: Medical Image Segmentation Using Grouped Coordinate Attention

Accurate segmentation of heterogeneous anatomical structures is pivotal for computer-aided diagnosis and subsequent clinical decision-making. Although U-Net based convolutional neural networks have achieved remarkable progress, their intrinsic locality and largely homogeneous attention formulations often limit the modeling of long-range contextual dependencies, especially in multi-organ scenarios and low-contrast regions. Transformer-based architectures mitigate this issue by leveraging global self-attention, but they usually require higher computational resources and larger training data, which may impede deployment in resource-constrained clinical environments.In this paper, we propose GCA-ResUNet, an efficient medical image segmentation framework equipped with a lightweight and plug-and-play Grouped Coordinate Attention (GCA) module. The proposed GCA decouples channel-wise context modeling into multiple groups to explicitly account for semantic heterogeneity across channels, and integrates direction-aware coordinate encoding to capture structured spatial dependencies along horizontal and vertical axes. This design enhances global representation capability while preserving the efficiency advantages of CNN backbones. Extensive experiments on two widely used benchmarks, Synapse and ACDC, demonstrate that GCA-ResUNet achieves Dice scores of 86.11% and 92.64%, respectively, outperforming a range of representative CNN and Transformer-based methods, including Swin-UNet and TransUNet. In particular, GCA-ResUNet yields consistent improvements in delineating small anatomical structures with complex boundaries. These results indicate that the proposed approach provides a favorable trade-off between segmentation accuracy and computational efficiency, offering a practical and scalable solution for clinical deployment.

cs.CV↗

arXiv2025

GCA-ResUNet:Image segmentation in medical images using grouped coordinate attention

Medical image segmentation underpins computer-aided diagnosis and therapy by supporting clinical diagnosis, preoperative planning, and disease monitoring. While U-Net style convolutional neural networks perform well due to their encoder-decoder structures with skip connections, they struggle to capture long-range dependencies. Transformer-based variants address global context but often require heavy computation and large training datasets. This paper proposes GCA-ResUNet, an efficient segmentation network that integrates Grouped Coordinate Attention (GCA) into ResNet-50 residual blocks. GCA uses grouped coordinate modeling to jointly encode global dependencies across channels and spatial locations, strengthening feature representation and boundary delineation while adding minimal parameter and FLOP overhead compared with self-attention. On the Synapse dataset, GCA-ResUNet achieves a Dice score of 86.11%, and on the ACDC dataset, it reaches 92.64%, surpassing several state-of-the-art baselines while maintaining fast inference and favorable computational efficiency. These results indicate that GCA offers a practical way to enhance convolutional architectures with global modeling capability, enabling high-accuracy and resource-efficient medical image segmentation.

cs.CV↗

arXiv2025

4D-PreNet: A Unified Preprocessing Framework for 4D-STEM Data Analysis

Automated experimentation with real time data analysis in scanning transmission electron microscopy (STEM) often require end-to-end framework. The four-dimensional scanning transmission electron microscopy (4D-STEM) with high-throughput data acquisition has been constrained by the critical bottleneck results from data preprocessing. Pervasive noise, beam center drift, and elliptical distortions during high-throughput acquisition inevitably corrupt diffraction patterns, systematically biasing quantitative measurements. Yet, conventional correction algorithms are often material-specific and fail to provide a robust, generalizable solution. In this work, we present 4D-PreNet, an end-to-end deep-learning pipeline that integrates attention-enhanced U-Net and ResNet architectures to simultaneously perform denoising, center correction, and elliptical distortion calibration. The network is trained on large, simulated datasets encompassing a wide range of noise levels, drift magnitudes, and distortion types, enabling it to generalize effectively to experimental data acquired under varying conditions. Quantitative evaluations demonstrate that our pipeline reduces mean squared error by up to 50% during denoising and achieves sub-pixel center localization in the center detection task, with average errors below 0.04 pixels. The outputs are bench-marked against traditional algorithms, highlighting improvements in both noise suppression and restoration of diffraction patterns, thereby facilitating high-throughput, reliable 4D-STEM real-time analysis for automated characterization.

cs.CV↗

arXiv2024

SUICA: Learning Super-high Dimensional Sparse Implicit Neural Representations for Spatial Transcriptomics

Spatial Transcriptomics (ST) is a method that captures gene expression profiles aligned with spatial coordinates. The discrete spatial distribution and the super-high dimensional sequencing results make ST data challenging to be modeled effectively. In this paper, we manage to model ST in a continuous and compact manner by the proposed tool, SUICA, empowered by the great approximation capability of Implicit Neural Representations (INRs) that can enhance both the spatial density and the gene expression. Concretely within the proposed SUICA, we incorporate a graph-augmented Autoencoder to effectively model the context information of the unstructured spots and provide informative embeddings that are structure-aware for spatial mapping. We also tackle the extremely skewed distribution in a regression-by-classification fashion and enforce classification-based loss functions for the optimization of SUICA. By extensive experiments of a wide range of common ST platforms under varying degradations, SUICA outperforms both conventional INR variants and SOTA methods regarding numerical fidelity, statistical correlation, and bio-conservation. The prediction by SUICA also showcases amplified gene signatures that enriches the bio-conservation of the raw data and benefits subsequent analysis. The code is available at https://github.com/Szym29/SUICA.

cs.LG↗

arXiv2024

Chalcogenide Metasurfaces Enabling Ultra-Wideband Detectors from Visible to Mid-infrared

Thermoelectric materials can be designed to support optical resonances across multiple spectral ranges to enable ultra-wide band photodetection. For instance, antimony telluride (Sb2Te3) chalcogenide exhibits interband plasmonic resonances in the visible range and Mie resonances in the mid-infrared (mid-IR) range, while simultaneously possessing large thermoelectric Seebeck coefficients. In this paper, we designed and fabricated Sb2Te3 metasurface devices to achieve resonant absorption for enabling photodetectors operating across an ultra-wideband spectrum, from visible to mid-IR. Furthermore, relying on asymmetric Sb2Te3 metasurface, we demonstrated the thermoelectric photodetectors with polarization-selectivity. This work provides a potential platform towards the portable ultrawide band spectrometers at room temperature, for environmental sensing applications.

physics.optics↗

arXiv2024

Predictive Modeling of Flexible EHD Pumps using Kolmogorov-Arnold Networks

We present a novel approach to predicting the pressure and flow rate of flexible electrohydrodynamic pumps using the Kolmogorov-Arnold Network. Inspired by the Kolmogorov-Arnold representation theorem, KAN replaces fixed activation functions with learnable spline-based activation functions, enabling it to approximate complex nonlinear functions more effectively than traditional models like Multi-Layer Perceptron and Random Forest. We evaluated KAN on a dataset of flexible EHD pump parameters and compared its performance against RF, and MLP models. KAN achieved superior predictive accuracy, with Mean Squared Errors of 12.186 and 0.001 for pressure and flow rate predictions, respectively. The symbolic formulas extracted from KAN provided insights into the nonlinear relationships between input parameters and pump performance. These findings demonstrate that KAN offers exceptional accuracy and interpretability, making it a promising alternative for predictive modeling in electrohydrodynamic pumping.

cs.LG↗

arXiv2023

scBeacon: single-cell biomarker extraction via identifying paired cell clusters across biological conditions with contrastive siamese networks

Despite the breakthroughs in biomarker discovery facilitated by differential gene analysis, challenges remain, particularly at the single-cell level. Traditional methodologies heavily rely on user-supplied cell annotations, focusing on individually expressed data, often neglecting the critical interactions between biological conditions, such as healthy versus diseased states. In response, here we introduce scBeacon, an innovative framework built upon a deep contrastive siamese network. scBeacon pioneers an unsupervised approach, adeptly identifying matched cell populations across varied conditions, enabling a refined differential gene analysis. By utilizing a VQ-VAE framework, a contrastive siamese network, and a greedy iterative strategy, scBeacon effectively pinpoints differential genes that hold potential as key biomarkers. Comprehensive evaluations on a diverse array of datasets validate scBeacon's superiority over existing single-cell differential gene analysis tools. Its precision and adaptability underscore its significant role in enhancing diagnostic accuracy in biomarker discovery. With the emphasis on the importance of biomarkers in diagnosis, scBeacon is positioned to be a pivotal asset in the evolution of personalized medicine and targeted treatments.

q-bio.GN↗

arXiv2023

Three-dimensional atomic positions and local chemical order of medium- and high-entropy alloys

Medium- and high-entropy alloys (M/HEAs) mix multiple principal elements with near-equiatomic composition and represent a paradigm-shift strategy for designing new materials for metallurgy, catalysis, and other fields. One of the core hypotheses of M/HEAs is lattice distortion. However, experimentally determining the 3D local lattice distortion in M/HEAs remains a challenge. Additionally, the presumed random elemental mixing in M/HEAs has been questioned by atomistic simulations, energy dispersive x-ray spectroscopy (EDS), and electron diffraction, which suggest the existence of local chemical order in M/HEAs. However, the 3D local chemical order has eluded direct experimental observation since the EDS elemental maps integrate the composition of atomic columns along the zone axes, and the diffuse reflections/streaks in electron diffraction of M/HEAs may originate from planar defects. Here, we determine the 3D atomic positions of M/HEA nanocrystals using atomic electron tomography, and quantitatively characterize the local lattice distortion, strain tensor, twin boundaries, dislocation cores, and chemical short-range order (CSRO) with unprecedented 3D detail. We find that the local lattice distortion and strain tensor in the HEAs are larger and more heterogeneous than in the MEAs. We observe CSRO-mediated twinning in the MEAs. that is, twinning occurs in energetically unfavoured CSRO regions but not in energetically favoured CSRO ones. This observation confirms the atomistic simulation results of the bulk CrCoNi MEA and represents the first experimental evidence of correlating local chemical order with structural defects in any material system. We expect that this work will not only expand our fundamental understanding of this important class of materials, but also could provide the foundation for tailoring M/HEA properties through lattice distortion and local chemical order.

cond-mat.mtrl-sci↗

arXiv2022

N\'eel Spin Currents in Antiferromagnets

Ferromagnets are known to support spin-polarized currents that control various spin-dependent transport phenomena useful for spintronics. On the contrary, fully compensated antiferromagnets are expected to support only globally spin-neutral currents. Here, we demonstrate that these globally spin-neutral currents can represent the N\'eel spin currents, i.e. staggered spin currents flowing through different magnetic sublattices. The N\'eel spin currents emerge in antiferromagnets with strong intra-sublattice coupling (hopping) and drive the spin-dependent transport phenomena such as tunneling magnetoresistance (TMR) and spin-transfer torque (STT) in antiferromagnetic tunnel junctions (AFMTJs). Using RuO$_{2}$ and Fe$_{4}$GeTe$_{2}$ as representative antiferromagnets, we predict that the N\'eel spin currents with a strong staggered spin-polarization produce a sizable field-like STT capable of the deterministic switching of the N\'eel vector in the associated AFMTJs. Our work uncovers the previously unexplored potential of fully compensated antiferromagnets and paves a new route to realize the efficient writing and reading of information for antiferromagnetic spintronics.

cond-mat.mes-hall↗

arXiv2022

Intrinsic shear transformations in metallic glasses

Plastic flow in amorphous solids is known to be carried by localized shear transformations (STs) which have been proposed to preferentially initiate from some defect units in the structure, akin to dislocations and point defects in crystalline solids. Despite the central role of STs in the mechanical deformation of metallic glasses (MGs), our knowledge of their characteristics has so far been limited to hypothetical models, based on computer simulations using unreleastically high cooling rates. Using combined molecular dynamics (MD) and Monte Carlo (MC) simulations, here we have succeeded in solidifying a metallic liquid at an effective cooling rate as slow as 500 K/s to approach that typical in experiments for producing bulk MGs. Exploiting this realistic MG model, we find that STs do not arise from signature structural defects that can be recognized a priori. Upon yielding, only about 2% of the total atoms participate in STs, each event involving as few as ~10 atoms. These findings rectify the unrealistically high content of liquid-like regions retained in MD-produced glass structures, which has rendered the MG model artificially ductile and under-predicted the sample-wide shear modulus by at least ~20% (with respect to that of experimental BMGs). Our finding sheds light on the scope of intrinsic structural inhomogeneity as well as the indeterministic aspect of the ST emergence under mechanical loading.

cond-mat.mtrl-sci↗

arXiv2022

Evidence of Spin Frustration in Vanadium Diselenide Monolayer Magnet

Monolayer VSe2, featuring both charge density wave and magnetism phenomena, represents a unique van der Waals magnet in the family of metallic two-dimensional transition-metal dichalcogenides (2D-TMDs). Herein, by means of in-situ microscopic and spectroscopic techniques, including scanning tunneling microscopy/spectroscopy, synchrotron X-ray and angle-resolved photoemission, and X-ray absorption, direct spectroscopic signatures are established, that identify the metallic 1T-phase and vanadium 3d1 electronic configuration in monolayer VSe2 grown on graphite by molecular-beam epitaxy. Element-specific X-ray magnetic circular dichroism, complemented with magnetic susceptibility measurements, further reveals monolayer VSe2 as a frustrated magnet, with its spins exhibiting subtle correlations, albeit in the absence of a long-range magnetic order down to 2 K and up to a 7 T magnetic field. This observation is attributed to the relative stability of the ferromagnetic and antiferromagnetic ground states, arising from its atomic-scale structural features, such as rotational disorders and edges. The results of this study extend the current understanding of metallic 2D-TMDs in the search for exotic low-dimensional quantum phenomena, and stimulate further theoretical and experimental studies on van der Waals monolayer magnets.

cond-mat.mtrl-sci↗

arXiv2021

Radiation-tolerant high-entropy alloys via interstitial-solute-induced chemical heterogeneities

High-entropy alloys (HEAs) composed of multiple principal elements have been shown to offer improved radiation resistance over their elemental or dilute-solution counterparts. Using NiCoFeCrMn HEA as a model, here we introduce carbon and nitrogen interstitial alloying elements to impart chemical heterogeneities in the form of the local chemical order (LCO) and associated compositional variations. Density functional theory simulations predict chemical short-range order (CSRO) (nearest neighbors and the next couple of atomic shells) surrounding C and N, due to the chemical affinity of C with (Co, Fe) and N with (Cr, Mn). Atomic-resolution chemical mapping of the elemental distribution confirms marked compositional variations well beyond statistical fluctuations. Ni+ irradiation experiments at elevated temperatures demonstrate a remarkable reduction in void swelling by at least one order of magnitude compared to the base HEA without C and N alloying. The underlying mechanism is that the interstitial-solute-induced chemical heterogeneities roughen the lattice as well as the energy landscape, impeding the movements of, and constraining the path lanes for, the normally fast-moving self-interstitials and their clusters. The irradiation-produced interstitials and vacancies therefore recombine more readily, delaying void formation. Our findings thus open a promising avenue towards highly radiation-tolerant alloys.

cond-mat.mtrl-sci↗

arXiv2020

Predicting the propensity for thermally activated $\beta$ events in metallic glasses via interpretable machine learning

The elementary excitations in metallic glasses (MGs), i.e., $\beta$ processes that involve hopping between nearby sub-basins, underlie many unusual properties of the amorphous alloys. A high-efficacy prediction of the propensity for those activated processes from solely the atomic positions, however, has remained a daunting challenge. Recently, employing well-designed site environment descriptors and machine learning (ML), notable progress has been made in predicting the propensity for stress-activated $\beta$ processes (i.e., shear transformations) from the static structure. However, the complex tensorial stress field and direction-dependent activation would induce non-trivial noises in the data, limiting the accuracy of the structure-property mapping learned. Here, we focus on the thermally activated elementary excitations and generate high-quality data in several Cu-Zr MGs, allowing quantitative mapping of the potential energy landscape. After fingerprinting the atomic environment with short- and medium-range interstice distribution, ML can identify the atoms with strong resistance or high compliance to thermal activation, at an unprecedented accuracy over ML models for stress-driven activation events. Interestingly, a quantitative "between-task" transferring test reveals that our learnt model can also generalize to predict the propensity of shear transformation. Our dataset is potentially useful for benchmarking future ML models on structure-property relationships in MGs.

cond-mat.mtrl-sci↗