arXiv subjects

Xi Luo

Publications and source records attributed to Xi Luo.

At least 19 recordsLinked to original sources

arXiv2026

HPSO: Particle Swarm Optimization with Hypergraph-Based Topology

Particle swarm optimization (PSO) has been widely applied to solve complex optimization problems from real-world applications due to its efficient exploration of large solution spaces and the ability to converge towards optimal solutions without requiring gradient information. Common swarm topologies in standard PSO and its variants, e.g., Ring and Star, can be regarded as graphs, where each edge connects only two particles. Such topology structures allow direct interactions only between connected particle pairs, and thus often fail to directly capture the higher-order social relationships that are necessary for navigating complex search landscapes. Therefore, this article proposes a novel PSO variant termed Hypergraph-assisted Particle Swarm Optimization (HPSO). In HPSO, the topology of the particles in a swarm is modeled by a hypergraph, in which hyperedges are used to connect multiple particles. This allows multiple particles within a hyperedge to interact directly. Furthermore, an adaptive hypergraph updating strategy is designed to periodically reconstruct the topology based on cumulative average particle displacement, thereby maintaining swarm diversity throughout the evolutionary process. In the experiments, the effectiveness of HPSO is verified on the IEEE CEC'17 benchmark suite, and the results demonstrate that HPSO achieves promising performance across various types of functions. Furthermore, the ablation experiment demonstrates that HPSO has excellent search capabilities.

cs.NE↗

arXiv2026

Monopole Spin Density Wave States in Magnetic Weyl Semimetals

The interplay between topology and magnetism in Weyl semimetals has recently emerged as a fertile ground for novel quantum phases. While monopole harmonic order parameters have been established for superconductivity and charge density waves in these systems, their spin density wave counterparts remain unexplored. Here we introduce monopole spin density wave (SDW) states arising from particle-hole pairing between nested Fermi surfaces enclosing Weyl nodes of the same chirality. We demonstrate that the SDW order parameter inherits a nontrivial pairing Berry phase and is described by monopole harmonic functions that exhibit topologically protected nodal structures in the gap function. Through a concrete lattice model, we show that helical and cycloidal SDW orders produce distinct signatures in band structures, Fermi arc distributions, and surface spin polarization patterns, which can be directly resolved by spin- and angle-resolved photoemission spectroscopy. Remarkably, we find that the quantum geometric tensor of the monopole pairing realizes ideal quantum geometry in the weak-coupling limit, where quantum distance fluctuations are entirely governed by Berry curvature. Our results not only unify the understanding of monopole ordered states across pairing channels but also provide experimental avenues for distinguishing competing magnetic orders in ReAlX (Re=rare earth elements, X=Si, Ge) materials and suggest potential applications in topological spintronics.

cond-mat.str-el↗

arXiv2026

PVDetector: Detecting Prompt Injection Attacks on Purpose-Specific LLM Agents through Policy-Violation Concept Analysis

Large language models (LLMs) are increasingly deployed as purpose-specific agents to handle domain-specific tasks such as customer service and code generation. These agents are expected to comply with not only generic safety guardrails but also purpose-specific restrictions tailored to their designated roles. Such additional restrictions enlarge the attack surface, particularly to prompt injection (PI) attacks. To defend against such attacks, existing detection methods primarily rely on analyzing input-output patterns, yet yield limited effectiveness. To address this limitation, we turn to analyzing the hidden activation space and discover that LLMs inherently retain latent policy-violation (PV) concepts when prompted with requests beyond their designated purpose. Particularly, PV concepts capture the semantics of conflicts between user queries and predefined restrictions, implicitly reflecting LLMs' intrinsic awareness of recognizing policy violations. Building on this insight, we propose PVDetector, a training-free framework that detects PI attacks during LLM inference by measuring hidden-state alignment with PV concepts, which are derived offline from the contrastive pairs of policy-violating and policy-compliant prompts. Experiments across multiple LLMs and datasets show that PVDetector achieves <1\% false negative rate with minimal auxiliary overhead, consistently outperforming state-of-the-art methods. Our code is available at https://github.com/Claresigle/PVDetector .

cs.CR↗

arXiv2026

Multilevel Regression Modeling of Covariance Matrix Outcomes

Covariance matrix outcomes arise naturally in neuroimaging experiments to study brain functional connectivity. It is also of interest to understand how brain network organization varies with subject-level covariates. Existing covariance regression methods operate in a single-level framework and do not accommodate the hierarchically nested data structure in which subjects are grouped into clusters, such as age cohorts in lifespan studies. A Multilevel Covariate-Assisted Principal Regression (MCAP) framework is introduced, which identifies, for each cluster, a linear projection such that a generalized linear mixed effects model can be formulated with the covariates. The cluster-specific projections are modeled on the unit sphere via a von Mises-Fisher distribution, enabling principled borrowing of information across clusters. Model parameters are estimated by maximizing a hierarchical likelihood. For inference, a two-stage bootstrap procedure is proposed. Asymptotic properties of the estimators are established. Simulation studies demonstrate that MCAP substantially outperforms single-level competitors in estimating regression coefficients. Applied to the Human Connectome Project Lifespan Study spanning ages from five to ninety, MCAP identifies a dominant spectral brain network capturing age and sex effects on functional connectivity, and reveals findings including the convergence of neural reorganization patterns in late adulthood and the coordinated lifespan modulation of cross-network regions linked to language and executive function.

stat.ME↗

arXiv2026

Skyrmion Phase and Non-Fermi Liquid Behavior in Nonsymmorphic Magnetic Weyl Semimetals

We investigate the interplay between complex magnetic orders and topological electronic states in nonsymmorphic magnetic Weyl semimetals of the ReAlX family (Re is a rare earth element and X is Si or Ge). We show that a Skyrmion lattice can fundamentally alter the behavior of Weyl fermions, driving the system into a non-Fermi liquid state and producing large, sign-tunable Hall responses. To this end, we construct a lattice model incorporating conduction Weyl fermions coupled to localized magnetic moments via Kondo interaction. Considering a multi-${\bf Q}$ cycloid magnetic configuration that evolves into a Skyrmion lattice under an in-plane Zeeman field, we analyze its profound impact on the band structure through magnetic Brillouin zone and band-folding. Using the Kubo formula, we calculate the conductivity tensor and examine the transport properties in the clean limit. Our results reveal that the Skyrmion lattice induces significant changes in both longitudinal and Hall conductivities. Remarkably, the temperature-dependent resistivity deviates from standard Fermi-liquid behavior ($\rho_{xx}\sim T^2$), exhibiting a non-Fermi liquid power-law scaling ($\rho_{xx}\sim T^\alpha$ with $\alpha$ between 3 and 5). This work provides a unified theoretical framework connecting multi-${\bf Q}$ magnetic textures, Skyrmion physics, and anomalous transport in topological semimetals, bridging the fields of topological magnetism and topological fermions.

cond-mat.mes-hall↗

arXiv2025

Graph Neural Networks with Transformer Fusion of Brain Connectivity Dynamics and Tabular Data for Forecasting Future Tobacco Use

Integrating non-Euclidean brain imaging data with Euclidean tabular data, such as clinical and demographic information, poses a substantial challenge for medical imaging analysis, particularly in forecasting future outcomes. While machine learning and deep learning techniques have been applied successfully to cross-sectional classification and prediction tasks, effectively forecasting outcomes in longitudinal imaging studies remains challenging. To address this challenge, we introduce a time-aware graph neural network model with transformer fusion (GNN-TF). This model flexibly integrates both tabular data and dynamic brain connectivity data, leveraging the temporal order of these variables within a coherent framework. By incorporating non-Euclidean and Euclidean sources of information from a longitudinal resting-state fMRI dataset from the National Consortium on Alcohol and Neurodevelopment in Adolescence (NCANDA), the GNN-TF enables a comprehensive analysis that captures critical aspects of longitudinal imaging data. Comparative analyses against a variety of established machine learning and deep learning models demonstrate that GNN-TF outperforms these state-of-the-art methods, delivering superior predictive accuracy for predicting future tobacco usage. The end-to-end, time-aware transformer fusion structure of the proposed GNN-TF model successfully integrates multiple data modalities and leverages temporal dynamics, making it a valuable analytic tool for functional brain imaging studies focused on clinical outcome prediction.

cs.LG↗

arXiv2025

TWLR: Text-Guided Weakly-Supervised Lesion Localization and Severity Regression for Explainable Diabetic Retinopathy Grading

Accurate medical image analysis can greatly assist clinical diagnosis, but its effectiveness relies on high-quality expert annotations Obtaining pixel-level labels for medical images, particularly fundus images, remains costly and time-consuming. Meanwhile, despite the success of deep learning in medical imaging, the lack of interpretability limits its clinical adoption. To address these challenges, we propose TWLR, a two-stage framework for interpretable diabetic retinopathy (DR) assessment. In the first stage, a vision-language model integrates domain-specific ophthalmological knowledge into text embeddings to jointly perform DR grading and lesion classification, effectively linking semantic medical concepts with visual features. The second stage introduces an iterative severity regression framework based on weakly-supervised semantic segmentation. Lesion saliency maps generated through iterative refinement direct a progressive inpainting mechanism that systematically eliminates pathological features, effectively downgrading disease severity toward healthier fundus appearances. Critically, this severity regression approach achieves dual benefits: accurate lesion localization without pixel-level supervision and providing an interpretable visualization of disease-to-healthy transformations. Experimental results on the FGADR, DDR, and a private dataset demonstrate that TWLR achieves competitive performance in both DR classification and lesion segmentation, offering a more explainable and annotation-efficient solution for automated retinal image analysis.

cs.CV↗

arXiv2025

Consistent gauge theories for the slave particle representation of the strongly correlated $t$-$J$ model

We aim to clarify the confusion and inconsistency in our recent works [1,2], and to address the incompleteness therein. In order to avoid the ill-defined nature of the free propagator of the gauge field in the ordered states of the $t$-$J$ model, we adopted a gauge fixing that was not of the Becchi-Rouet-Stora-Tyutin (BRST) exact form in our previous work [2]. This led to the situation where Dirac's second-class constraints, namely, the slave particle number constraint and the Ioffe-Larkin current constraint, were not rigorously obeyed. Here we show that a consistent gauge fixing condition that enforces the exact constraints is BRST-exact in our theory. An example is the Lorenz gauge. On the other hand, we prove that although the free propagator of the gauge field in the Lorenz gauge is ill-defined, the full propagator is still well-defined. This implies that the strongly correlated $t$-$J$ model can be exactly mapped to a perturbatively controllable theory within the slave particle representation.

cond-mat.str-el↗

arXiv2025

Predictiveness Curve Assessment under Competing Risks for Risk Prediction Models

The predictiveness curve is a valuable tool for predictive evaluation, risk stratification, and threshold selection in a target population, given a single biomarker or a prediction model. In the presence of competing risks, regression models are often used to generate predictive risk scores or probabilistic predictions targeting the cumulative incidence function--distinct from the cumulative distribution function used in conventional predictiveness curve analyses. We propose estimation and inference procedures for the predictiveness curve with a competing risks regression model, to display the relationship between the cumulative incidence probability and the quantiles of model-based predictions. The estimation procedure combines cross-validation with a flexible regression model for tau-year event risk given the model-based risk score, with corresponding inference procedures via perturbation resampling. The proposed methods perform satisfactorily in simulation studies and are implemented through an R package. We apply the proposed methods to a cirrhosis study to depict the predictiveness curve with model-based predictions for liver-related mortality.

stat.ME↗

arXiv2025

TrioXpert: An Automated Incident Management Framework for Microservice System

Automated incident management plays a pivotal role in large-scale microservice systems. However, many existing methods rely solely on single-modal data (e.g., metrics, logs, and traces) and struggle to simultaneously address multiple downstream tasks, including anomaly detection (AD), failure triage (FT), and root cause localization (RCL). Moreover, the lack of clear reasoning evidence in current techniques often leads to insufficient interpretability. To address these limitations, we propose TrioXpert, an end-to-end incident management framework capable of fully leveraging multimodal data. TrioXpert designs three independent data processing pipelines based on the inherent characteristics of different modalities, comprehensively characterizing the operational status of microservice systems from both numerical and textual dimensions. It employs a collaborative reasoning mechanism using large language models (LLMs) to simultaneously handle multiple tasks while providing clear reasoning evidence to ensure strong interpretability. We conducted extensive evaluations on two microservice system datasets, and the experimental results demonstrate that TrioXpert achieves outstanding performance in AD (improving by 4.7% to 57.7%), FT (improving by 2.1% to 40.6%), and RCL (improving by 1.6% to 163.1%) tasks. TrioXpert has also been deployed in Lenovo's production environment, demonstrating substantial gains in diagnostic efficiency and accuracy.

cs.SE↗

arXiv2025

Correlated helimagnetic configuration in a nonsymmorphic magnetic nodal semimetal

Nonsymmorphic magnetic Weyl semimetal materials such as ReAlX (Re=rare earth, X=Si/Ge) provide a unique opportunity to explore the correlated phenomena between Weyl fermions and nontrivial magnetic configurations. To be specific, we study a lattice model in which the magnetic configuration is determined by the competition among ferromagnetic (FM) interaction, the Dzyaloshinskii-Moriya interaction, and the Kondo coupling $K_0$ to the Weyl fermion. Both quantum and finite-temperature phase transitions between FM and correlated nesting helical configurations are found. Different from the uncorrelated helimagnet that decouples from the Weyl fermions, this correlated helimagnet induces a magnetic Brillouin zone with a $K_0$-dependent nesting in the band structure of the conduction electrons instead of the monopole-like Weyl cone. By measuring the current induced by the chiral magnetic effect on the conduction electron with nesting Weyl nodes, one can distinguish the correlated nesting helical order from the ferromagnetism because the chiral magnetic effect is considerably suppressed in the former case. These properties we find here may explain the experimental observations in ReAlX.

cond-mat.mes-hall↗

arXiv2025

Knowledge Distillation with Adapted Weight

Although large models have shown a strong capacity to solve large-scale problems in many areas including natural language and computer vision, their voluminous parameters are hard to deploy in a real-time system due to computational and energy constraints. Addressing this, knowledge distillation through Teacher-Student architecture offers a sustainable pathway to compress the knowledge of large models into more manageable sizes without significantly compromising performance. To enhance the robustness and interpretability of this framework, it is critical to understand how individual training data impact model performance, which is an area that remains underexplored. We propose the \textbf{Knowledge Distillation with Adaptive Influence Weight (KD-AIF)} framework which leverages influence functions from robust statistics to assign weights to training data, grounded in the four key SAFE principles: Sustainability, Accuracy, Fairness, and Explainability. This novel approach not only optimizes distillation but also increases transparency by revealing the significance of different data. The exploration of various update mechanisms within the KD-AIF framework further elucidates its potential to significantly improve learning efficiency and generalization in student models, marking a step toward more explainable and deployable Large Models. KD-AIF is effective in knowledge distillation while also showing exceptional performance in semi-supervised learning with outperforms existing baselines and methods in multiple benchmarks (CIFAR-100, CIFAR-10-4k, SVHN-1k, and GLUE).

cs.LG↗

arXiv2025

Tracing Partisan Bias to Its Emotional Fingerprints: A Computational Approach to Mitigation

This study introduces a novel framework for analysing and mitigating media bias by tracing partisan stances to their linguistic roots in emotional language. We posit that partisan bias is not merely an abstract stance but materialises as quantifiable 'emotional fingerprints' within news texts. These fingerprints are systematically measured using the Valence-Arousal-Dominance (VAD) framework, allowing us to decode the affective strategies behind partisan framing. Our analysis of the Allsides dataset confirms this hypothesis, revealing distinct and statistically significant emotional fingerprints for left, centre, and right-leaning media. Based on this evidence-driven approach, we then propose a computational approach to mitigation through NeutraSum, a model designed to neutralise these identified emotional patterns. By explicitly targeting the VAD characteristics of biased language, NeutraSum generates summaries that are not only coherent but also demonstrably closer to an emotionally neutral baseline. Experimental results validate our framework: NeutraSum successfully erases the partisan emotional fingerprints from its summaries, achieving a demonstrably lower emotional bias score than other models. This work pioneers a new path for bias mitigation, shifting the focus from treating symptoms (political labels) to addressing the cause: the emotional encoding of partisan bias in language.

cs.CL↗

arXiv2024

Air-Ground Collaborative Robots for Fire and Rescue Missions: Towards Mapping and Navigation Perspective

Air-ground collaborative robots have shown great potential in the field of fire and rescue, which can quickly respond to rescue needs and improve the efficiency of task execution. Mapping and navigation, as the key foundation for air-ground collaborative robots to achieve efficient task execution, have attracted a great deal of attention. This growing interest in collaborative robot mapping and navigation is conducive to improving the intelligence of fire and rescue task execution, but there has been no comprehensive investigation of this field to highlight their strengths. In this paper, we present a systematic review of the ground-to-ground cooperative robots for fire and rescue from a new perspective of mapping and navigation. First, an air-ground collaborative robots framework for fire and rescue missions based on unmanned aerial vehicle (UAV) mapping and unmanned ground vehicle (UGV) navigation is introduced. Then, the research progress of mapping and navigation under this framework is systematically summarized, including UAV mapping, UAV/UGV co-localization, and UGV navigation, with their main achievements and limitations. Based on the needs of fire and rescue missions, the collaborative robots with different numbers of UAVs and UGVs are classified, and their practicality in fire and rescue tasks is elaborated, with a focus on the discussion of their merits and demerits. In addition, the application examples of air-ground collaborative robots in various firefighting and rescue scenarios are given. Finally, this paper emphasizes the current challenges and potential research opportunities, rounding up references for practitioners and researchers willing to engage in this vibrant area of air-ground collaborative robots.

cs.RO↗

arXiv2024

Deep Learning the Forecast of Galactic Cosmic-Ray Spectra

We introduce a novel deep learning framework based on Long Short-Term Memory (LSTM) networks to predict galactic cosmic-ray spectra on a one-day-ahead basis by leveraging historical solar activity data, overcoming limitations inherent in traditional transport models. By flexibly incorporating multiple solar parameters, such as the heliospheric magnetic field, solar wind speed, and sunspot numbers, our model achieves accurate short-term and long-term predictions of cosmic-ray flux. The addition of historical cosmic-ray flux data significantly enhances prediction accuracy, allowing the model to capture complex dependencies between past and future flux variations. Additionally, the model reliably predicts full cosmic-ray spectra for different particle species, enhancing its utility for comprehensive space weather forecasting. Our approach offers a scalable, data-driven alternative to traditional physics-based methods, ensuring robust daily and long-term forecasts. This work opens avenues for advanced models that can integrate broader observational data, with significant implications for space weather monitoring and mission planning.

astro-ph.HE↗

arXiv2024

Dissipationless topological quantum computation for Majorana objects in sparse-dense mixed encoding process

Topological quantum computation based on Majorana objects is subject to a significant challenge because at least some of the two-qubit quantum gates rely on the fermion (either charge or spin) parity of the qubits. This dependency renders the quantum operations involving these gates probabilistic when attempting to advance quantum processes within the quantum circuit model. Such an approach leads to significant information loss whenever measurements yield the undesired fermion parity. To resolve the problem of wasting information, we devise topological operations that allow for the non-dissipative correction of information from undesired fermion parity to the desired one. We will use the sparse-dense mixed encoding process for the controlled-NOT gate as an example to explain how corrections can be implemented without affecting the quantum information carried by the computational qubits. This correction process can be applied {to} either the undesired input qubits or the fermion parity-dependent quantum gates, and it works for both Majorana-zero-mode-based and Majorana-edge-mode-based topological quantum computation.

quant-ph↗

arXiv2024

Online Ecological Gearshift Strategy via Neural Network with Soft-Argmax Operator

This paper presents a neural network optimizer with soft-argmax operator to achieve an ecological gearshift strategy in real-time. The strategy is reformulated as the mixed-integer model predictive control (MIMPC) problem to minimize energy consumption. Then the outer convexification is introduced to transform integer variables into relaxed binary controls. To approximate binary solutions properly within training, the soft-argmax operator is applied to the neural network with the fact that all the operations of this scheme are differentiable. Moreover, this operator can help push the relaxed binary variables close to 0 or 1. To evaluate the strategy effect, we deployed it to a 2-speed electric vehicle (EV). In contrast to the mature solver Bonmin, our proposed method not only achieves similar energy-saving effects but also significantly reduces the solution time to meet real-time requirements. This results in a notable energy savings of 6.02% compared to the rule-based method.

eess.SY↗

arXiv2023

Intrinsic chiral topological superconductor thin films

Superconductors (SCs) with nontrivial topological band structures in the normal state have been discovered recently in bulk materials. When such SCs are made into thin films, quantum tunneling and Cooper pairing take place between the topological surface states (TSSs) on the opposing surfaces. Here, we find that chiral topological superconductivity with spontaneous time-reversal symmetry breaking emerges on the surface of such thin film SCs. There is a mirror symmetry that protects a novel nonunitary orbital and spin triplet pairing of the TSS. In the mirror diagonal space, the chiral topological SC manifests as two independent chiral $p$-wave spin-triplet pairing states, in which each is a two-dimensional superconducting analog of the Anderson-Brinkman-Morel state in superfluid $^3$He with in-plane exchange fields. A rich topological phase diagram governed by the nontrivial $\mathbb{Z}\oplus\mathbb{Z}$ topological invariant is obtained with gapless chiral Majorana edge modes and anyonic Majorana vortices. We further construct a three dimensional lattice model with a topological band structure and SC pairings, which is motivated by Fe-based SCs such as Fe(Te,Se). We demonstrate the realization of the proposed intrinsic chiral topological superconductor in the quasi-two-dimensional thin-film limit. Our findings enable thin-film SCs with nontrivial $Z_2$ band structures as a single-material platform for intrinsic chiral topological superconductivity with both vortex and boundary Majorana excitations for topological quantum device making.

cond-mat.supr-con↗