arXiv subjects

Wei Sun

Publications and source records attributed to Wei Sun.

At least 37 records · Page 2Linked to original sources

arXiv2026

Scalar glueball-$s\bar{s}$ mixing in one flavor lattice QCD

We investigate the mixing between the lowest-lying scalar glueball and the $s\overline{s}$ meson in $N_f=1$ lattice quantum chromodynamics (QCD) utilizing an anisotropic $16^3 \times 128$ lattice ensemble at a lattice spacing $a_s\approx 0.148\,\rm{fm}$. By solving a generalized eigenvalue problem (GEVP) for the optimized glueball and $s\overline{s}$ scalar operators in the $J^\text{PC} = 0^{++}$ channel, the masses of the two lowest-lying eigenstates are determined to be $m_1 = 1.290(20)\,\mathrm{GeV}$ and $m_2 = 1.777(49)\,\mathrm{GeV}$. By extracting the couplings of these mass eigenstates to the glueball and $s\bar{s}$ operators, we determine a substantial mixing angle $|θ| \approx 40.7(2.7)^\circ$ and a large mixing energy $x_s = 239(24)$ MeV. These results indicate a strong glueball-$s\bar{s}$ mixing in the scalar sector, providing important non-perturbative inputs for understanding the nature of the experimental isoscalar scalar mesons. The continuum limit of the mixing energy and its quark mass dependence need to be investigated in the future.

hep-lat↗

arXiv2026

SafeAtlas-VL: Beyond Binary Multimodal Safety with Large-Scale Data and Guard Models

Multimodal safety moderation requires distinguishing risks arising from visual content, user intent, and assistant behavior. Existing safeguards, however, are typically trained for a single judgment target and reduce safety assessment to a binary decision. Consequently, risk becomes difficult to compare across a multimodal interaction, and ambiguous cases are obscured. We introduce SafeAtlas-VL, a dataset of 1.5M training instances that places image-, request-, and response-level judgments on a five-level ordered scale. We curate a broad collection of safety-relevant data from both real-world and synthetic sources and apply a disagreement-aware annotation procedure. The resulting dataset spans 15 harm categories and 55 fine-grained subcategories, covering a broad range of multimodal safety scenarios. We also construct SafeAtlas-Bench, a held-out set of 5,000 instances for evaluating five-level predictions and continuous risk scores. Upon this dataset, we train the SafeAtlas Guard series of models via target-conditioned tuning for multimodal safety detection. Our models not only perform five-way classification of safety levels but also map safety to continuous scores through a soft cumulative ordinal head. Experimental results demonstrate that guard models trained on our dataset exhibit strong generalization: even without using the training sets of other benchmarks, they achieve competitive performance on the corresponding test sets. Notably, our 8B model attains the overall best performance, outperforming the previous SOTA by approximately 4% in F1 score. Code, data, and models are released to support further research. Warning: this paper contains example data that may be offensive, harmful, graphic, or disturbing.

cs.AI↗

arXiv2026

Results related to the Gaussian product inequality conjecture for mixed-sign exponents in arbitrary dimension

This paper studies Gaussian product inequalities (GPIs) for centered Gaussian random vectors when positive and negative exponents appear simultaneously. We prove a quantitative lower bound for arbitrary mixed-sign patterns, conditional on the validity of a lower-dimensional GPI, and an upper bound for products of negative powers with convex functions of the remaining Gaussian components. The latter yields an explicit upper bound for mixed-sign products whenever the total positive-side exponent is at least $1/2$.

math.PR↗

arXiv2026

Scalar and tensor structures in $J/ψJ/ψ$ scattering from lattice QCD

The $^1S_0$ and $^5S_2$ amplitudes of $J/ψ~J/ψ$ scattering are determined up to 6.6~GeV from $N_f=2$ lattice QCD simulations at $m_π\approx 420$~MeV and $250$~MeV. We observe an attractive interaction in the ${}^1S_0$ channel, which permits the existence of a near-threshold scalar structure. This structure may correspond to $X(6200)$, but the possible left-hand-cut effect prevents us from a precise pole determination. In the ${}^5S_2$ channel, while a near-threshold repulsive interaction is observed, a resonance appears at higher energies associated with a Castillejo-Dalitz-Dyson zero $\sqrt{s}_{\rm CDD}\approx 6.45~\rm{GeV}$ of the scattering amplitude. The tensor resonance has the parameters $(m_R,Γ_R)=(6.543(10), 0.548(34))~\rm{GeV}$ for $m_π\approx 420~\rm{MeV}$ and $(6.538(13),0.537(56))~\rm{GeV}$ for $m_π\approx 250~\rm{MeV}$, which are compatible with those of $X(6600)$ (or $X(6400)$) reported by ATLAS and CMS, and also support the $2^{++}$ assignment of the recent CMS study. The difference of the $^1S_0$ and $^5S_2$ $J/ψJ/ψ$ interaction is attributed to the dominance of the quark rearrangement effects. The systematic uncertainties owing to the unphysical pion mass, the finite lattice spacing, and the finite volume should be investigated by future lattice QCD studies with more sophisticated lattice setups.

hep-lat↗

arXiv2026

Cross-lingual Representation Learning via Centroid Intervention Fusion

Large language models (LLMs) exhibit uneven multilingual performance, especially when dealing with low-resource languages. Inference-time intervention offers a lightweight way to improve cross-lingual transfer by modifying the hidden states produced by the LLMs during the forward pass, without updating model parameters. However, existing cross-lingual intervention methods typically learn separate projections from source to target languages, which limits scalability and prevents knowledge sharing across languages. We propose Centroid Intervention Fusion (CIF), a projection fusion framework that consolidates multiple multilingual intervention projections into a single language-shared operator. Across multilingual commonsense reasoning, natural language inference, factual editing, and machine translation benchmarks, CIF outperforms the strongest prior pairwise intervention baseline by up to +3.378 pp on average across four model backbones, while supporting performance gains for low resource languages. The code is available at https://github.com/VRCMF/CIF.git.

cs.CL↗

arXiv2026

ArmorOCR: Grounded Adversarial Visual Perception via Observation-Transferred Self-Distillation

Large multimodal models (LMMs) have demonstrated strong OCR recognition capabilities, yet remain vulnerable to adversarial visual text that is readable to humans but challenging for models to localize and recognize. Existing OCR benchmarks mainly focus on natural or document-style text, while adversarial OCR evaluations remain limited in scale, task coverage, or region-aware evaluation. In this paper, we formulate adversarial OCR as a \textbf{grounded OCR perception} task and introduce \textbf{AdvSpot}, the first benchmark for grounded adversarial OCR evaluation. AdvSpot comprises 390 images with region-level annotations, spanning 5 primary categories and 13 fine-grained adversarial OCR types. To address this challenge, we propose \textbf{ArmorOCR}, a two-stage training framework for robust adversarial OCR perception. ArmorOCR first acquires missing adversarial OCR perception from privileged transformed observations through On-Policy Self-Distillation (OPSD), and then refines grounded OCR perception through Group Relative Policy Optimization (GRPO) with task-conditioned rewards for localization, recognition, full spotting, and visual question answering (VQA). Experiments on our AdvSpot, other adversarial OCR benchmarks, and general OCR benchmarks demonstrate that ArmorOCR consistently improves adversarial OCR perception while preserving competitive general OCR capability.

cs.CV↗

arXiv2026

$η_cη_c$ and $J/ψJ/ψ$ scatterings from lattice QCD

We investigate the $S$-wave $η_cη_c$ and $J/ψJ/ψ$ scattering in the $J^{PC}=(0,2)^{++}$ channels up to a center-of-mass energy of 6.6~GeV. The calculations are carried out at two unphysical pion masses, $m_π\approx 420$~MeV and 250~MeV, in $N_f=2$ lattice QCD. For each $m_π$, we extract the finite-volume energy levels on two lattices with an identical lattice spacing ($a\simeq 0.136~\mathrm{fm}$) but different spatial volumes. Since the coupled-channel effects between the $η_cη_c$ and $J/ψJ/ψ$ channels are found to be negligible, we analyze the corresponding scattering properties using the single-channel L"{u}scher method. We find that the interactions in these dicharmonium systems are dominated by the quark rearrangement effect. In the $0^{++}$ channel, the near-threshold attraction in $J/ψJ/ψ$ and repulsion in $η_cη_c$ can be explained through the Fierz rearrangement. The attractive interaction in the ${}^1S_0$ $J/ψJ/ψ$ channel allows for the existence of a near-threshold scalar structure, which may correspond to the $X(6200)$. However, possible left-hand-cut effects due to light-hadron exchange prevent us from precisely determining the pole positions. In the $2^{++}$ channel, while the ${}^5S_2$ $J/ψJ/ψ$ system exhibits a repulsive interaction near threshold, the scattering amplitude has a Castillejo-Dalitz-Dyson zero at $\sqrt{s}=6.45~\mathrm{GeV}$ and a resonance pole at $\sqrt{s}=\big[6.543(10)-i,0.548(34)/2\big]~\mathrm{GeV}$ for $m_π\approx 420~\mathrm{MeV}$ and $\big[6.538(13)-i,0.537(56)/2\big]~\mathrm{GeV}$ for $m_π\approx 250~\mathrm{MeV}$, where the uncertainties are statistical. This resonance may correspond to the $X(6600)$ (or $X(6400)$) reported by the ATLAS and CMS Collaborations. Our result supports its $2^{++}$ assignment, in agreement with the latest spin-parity determination by CMS.

hep-lat↗

arXiv2026

Exploring Instruction Data Quality for Explainable Image Quality Assessment

In recent years, with the rapid development of large multimodal models (LMMs), explainable image quality assessment (IQA) has attracted increasing attention, aiming to understand the perceptual quality problems of images. Existing studies typically construct large-scale instruction tuning datasets to enhance the quality perception capabilities of LMMs, following the data scaling law. However, as the fundamental capabilities of LMMs continue to improve, existing instruction tuning datasets may contain redundant and less challenging samples, resulting in substantial computational costs. In this paper, we investigate whether data quantity remains the dominant factor in LMM instruction tuning for explainable IQA. Based on a strong pre-trained LMM, we observe that randomly selecting a properly sized subset of training data can outperform full-data fine-tuning, indicating substantial redundancy in existing instruction tuning datasets. Motivated by this observation, we propose Q-Selector, a clustering-based data selection framework consisting of three stages: hierarchical LMM-based clustering feature extraction, cluster quota allocation through density and transferability, and SVD-based cluster sampling strategy. Specifically, Q-Selector extracts multi-layer features from the vision encoder, text encoder, and large language model components of LMMs as clustering features. It then determines the sampling quota for each cluster based on inter-cluster transferability and intra-cluster density. Finally, Q-Selector employs a Singular Value Decomposition (SVD)-based sampling strategy to select high-quality instruction data. Experimental results demonstrate that Q-Selector achieves 102.1% and 103.7% of the performance of full-data fine-tuning using only 10% of the training data on explainable IQA and image aesthetics assessment tasks, respectively.

cs.CV↗

arXiv2026

Technical Report on Resilient and Secure Large-Scale Energy Internet Systems

This IEEE PES Task Force report examines the security and resilience of large-scale Energy Internet (EI) systems, in which electricity, information, and market layers are tightly coupled through pervasive digitalization. The report characterizes the EI cyber-physical threat landscape and surveys detection, assurance, and mitigation techniques, presents modeling, control, and decision-making frameworks that capture cyber-physical interdependencies, including storage integration, multi-dimensional resilience, and electricity price forecasting, examines adversarial risks and trustworthy deployment of artificial intelligence, and introduces graph-based, attack-resilient information routing. The report closes with recommendations for research, standardization, and regulatory efforts needed to realize a resilient and secure large-scale EI.

eess.SY↗

arXiv2026

Quarkyonic matter with strangeness in an extended relativistic mean-field model

Quarkyonic matter is expected to play a key role for the transition from hadronic matter to quark matter in compact stars. Within the framework of the relativistic mean-field (RMF) model and equivparticle model with density-dependent quark masses, we construct the ``quark Fermi sea" with a ``baryon Fermi surface" to characterize the properties of the quarkyonic matter. In particular, we develop a comprehensive framework to account for the strangeness degrees of freedom, incorporating $Λ$, $Ξ$, and $Σ$ hyperons as well as strange quarks in a unified quarkyonic framework. Our calculations indicate that the inevitable emergence of hyperons softens the equations of state, leading to a reduction in the equilibrium sound velocity around $n_{\rm b}\approx 2n_0$, and consequently reducing the masses and radii of neutron stars. When the quark-hadron phase transition is taken into account, the equation of state at high densities exhibits additional softening consistent with current astronomical observational constraints. This softening leads to a maximum equilibrium sound velocity of $v_{eq}^{\rm max} \approx 0.6\,c$, which is close to the ultrarelativistic limit of $0.58\,c$.

nucl-th↗

arXiv2026

Do LLMs Benefit From Their Own Words?

In multi-turn conversations, large language models typically condition on the full conversation history: both past user prompts and assistant responses. We revisit this design choice by comparing full-context prompting to four alternative, substantially-reduced context configurations. Analyzing in-the-wild multi-turn conversations across three open reasoning and one state-of-the-art model, we find that response quality is largely preserved under aggressive context filtering: replacing all prior assistant turns with one-sentence summaries or keeping only the most recent user--assistant exchange often matches storing full context in performance while using roughly 8x less context. To understand this result, we observe that a substantial fraction of user turns (36.4%) in multi-turn conversations are self-contained and that many follow-up turns can be addressed by seeing only the immediately preceding user--assistant exchange. Furthermore, we find that when models condition on their own past responses, this can lead to context pollution, a phenomenon in which reasoning errors, hallucinations, or stylistic artifacts propagate across turns. Motivated by these findings, we design a context-filtering approach that selectively omits the assistant-side history. Taken together, these findings suggest moving away from storing full dialogue transcripts and instead retaining only what is relevant.

cs.CL↗

arXiv2026

LoViF 2026 Challenge on Human-oriented Semantic Image Quality Assessment: Methods and Results

This paper reviews the LoViF 2026 Challenge on Human-oriented Semantic Image Quality Assessment. This challenge aims to raise a new direction, i.e., how to evaluate the loss of semantic information from the human perspective, intending to promote the development of some new directions, like semantic coding, processing, and semantic-oriented optimization, etc. Unlike existing datasets of quality assessment, we form a dataset of human-oriented semantic quality assessment, termed the SeIQA dataset. This dataset is divided into three parts for this competition: (i) training data: 510 pairs of degraded images and their corresponding ground truth references; (ii) validation data: 80 pairs of degraded images and their corresponding ground-truth references; (iii) testing data: 160 pairs of degraded images and their corresponding ground-truth references. The primary objective of this challenge is to establish a new and powerful benchmark for human-oriented semantic image quality assessment. There are a total of 58 teams registered in this competition, and 6 teams submitted valid solutions and fact sheets for the final testing phase. These submissions achieved state-of-the-art (SOTA) performance on the SeIQA dataset.

cs.CV↗

arXiv2026

Néel-Vector-Orientation Induced Direction-Robust Spin Filtering in Two-Dimensional Altermagnets

Whether an antiferromagnet can host direction-robust spin-polarized transport without a conventional spin-selective band gap remains a central challenge in antiferromagnetic spintronics. Here we establish a gapless, direction-robust spin-filtering mechanism in a compensated two-dimensional altermagnetic Weyl semimetal that requires neither a spin-selective band gap nor a large velocity contrast between spin projections. Using Janus monolayer Ta$_2$TeSeO as a realistic platform, we combine symmetry analysis with first-principles calculations, full-Brillouin-zone Wannier interpolation, and semiclassical transport. Rotating the Néel vector removes a unitary-mirror constraint and shifts one Weyl-cone pair away from its parent high-symmetry line. For an in-plane Néel vector, the residual $C_{2z}\mathcal T$ symmetry forbids the independent $σ_y$ mass that would open a local gap, allowing the reconstructed cones to shift in momentum while remaining gapless. Breaking unitary $C_{2z}$ simultaneously lifts the energy equivalence of the remaining mirror-pinned Weyl cones. The resulting coexistence of a metallic spin-projected manifold and a low-DOS Weyl-derived manifold produces a predominantly DOS-driven conductance imbalance. At charge neutrality and 20~K, the longitudinal conductivity polarization for $\mathbf n\parallel x$ remains positive for every in-plane current direction and ranges from $76.4\%$ to $82.0\%$. The degenerate in-plane magnetic anisotropy facilitates reversible switching between symmetry-related spin-filtering states using strain or weak anisotropic fields. This Néel-vector-driven symmetry mechanism provides a general route to direction-robust gapless spin filtering in compensated altermagnets.

cond-mat.mes-hall↗

arXiv2026

A Structured Knowledge Infrastructure for Domain-Specific Data Asset Discovery

Enterprise data analytics agents face two structural failures: generic RAG retrieves the wrong asset (Hit@10=19.1%) and delivers no usage knowledge to prevent metric misinterpretation---stemming from four root causes (C1--C4) ranging from semantic gap and entity ambiguity to schema drift and asset-usage gap. We present a two-layer solution deployed in the commercial advertising data warehouse at Xiaohongshu (5,300+ Hive tables, 14 domains). A three-tier dual-purpose knowledge base (179 documents, eight-section annotation template) serves both retrieval and generation, with a closed-loop refresh pipeline maintaining day-level freshness (one yes/no approval, 30s hot-reload). The Graph-Guided Retriever (GGR) uses a 2,859-node knowledge graph as a candidate gate with intent routing to deliver 71.6x token reduction. The Scene-Aware Ranker (SAR) applies 19-class entity recognition and explicit scenario annotations; negative knowledge alone contributes 25 percentage points of Hit@10 gain. On two 100-question benchmarks, Hit@10 rises from 19.1% to 96.6% (+77.5pp) and knowledge coverage from 56% to 77%, at 4.84--5.33s end-to-end latency.

cs.IR↗

arXiv2026

Counterfactual Optimal Action Trees (COAT): Interpretable Prescriptive Policies from Observational Data

We introduce COAT (Counterfactual Optimal Action Tree), a framework for learning interpretable prescriptive policies from observational data. COAT combines counterfactual outcome estimation with large-scale mixed-integer optimization, using column generation to translate causal predictions into feasible, transparent decisions under business and regulatory constraints. We apply COAT to airline ancillary pricing, a setting characterized by complex business rules and limited experimental flexibility. In a 17-week field pilot with a major global airline, COAT increased upsell revenue per booking by 6.9%, with the airline projecting \$50-\$150 million in incremental annual premium seat revenue across eligible domestic markets. The success of the pilot led to scaled adoption and informed broader AI-driven decision initiatives within the organization.

cs.LG↗

arXiv2026

CompressedVQA-AEV: Full-Reference and No-Reference Quality Assessment Models for Asymmetric Encoded Videos

This report presents our solutions to the QoMEX 2026 Grand Challenge on Video Quality Assessment for Asymmetric Encoded Videos, comprising a full-reference (FR) model, CompressedVQA-AEV-FR, and a no-reference (NR) model, CompressedVQA-AEV-NR. The FR approach leverages a Swin-B backbone to extract multi-stage similarity statistics between reference and distorted videos for quality prediction. For the NR setting, our model employs complementary frame-level encoders based on SigLIP2 and Swin-B, followed by temporal mean pooling and cross-fold ensembling to estimate perceptual quality without reference data. Our CompressedVQA-AEV-FR achieves first place in the FR track of QoMEX 2026 Grand Challenge, while CompressedVQA-AEV-NR secures fourth place in the NR track, demonstrating the effectiveness of our proposed models. The code is available at https://github.com/sunwei925/CompressedVQA-AEV.

eess.IV↗

arXiv2026

ICME 2026 Grand Challenge on Cross-Scenario Defect Detection and Fine-Grained Severity Grading for High-Precision Manufacturing

This paper presents the IEEE International Conference on Multimedia and Expo (ICME) 2026 Grand Challenge on Cross-Scenario Defect Detection and Fine-Grained Severity Grading for High-Precision Manufacturing. The challenge is motivated by two key limitations of existing industrial defect inspection systems: (1) current deep learning-based methods often suffer significant performance degradation when deployed in unseen production scenarios, and (2) most benchmarks neglect severity-aware assessment, which is critical for risk control and yield optimization. To address these limitations, we design two complementary tracks: Track 1 (Cross-Scenario Defect Detection) targets accurate defect detection, localization, and classification across diverse unseen production environments; Track 2 (Fine-Grained Severity Grading) requires assigning each detected defect an industry-standard severity level, including Acceptable, Marginal NG, NG, and Gross NG. We construct a large-scale industrial dataset of high-resolution microscopic images spanning seven representative defect categories, comprising over 3,800 images with pixel-level instance annotations for Track 1 and over 2,600 images with severity-grade labels for Track 2. The challenge attracted 86 registered participants with 130 submissions; during the final testing phase, 21 teams submitted results and 12 teams provided models with technical reports. The resulting benchmark, together with the diverse and effective solutions contributed by participating teams, sets a new standard for industrial defect analysis research.

cs.CV↗

arXiv2026

Occlusion-Aware 3D Hand-Object Pose Estimation with Masked AutoEncoders

Hand-object pose estimation from monocular RGB images remains a significant challenge mainly due to the severe occlusions inherent in hand-object interactions. Existing methods do not sufficiently explore global structural perception and reasoning, which limits their effectiveness in handling occluded hand-object interactions. To address this challenge, we propose an occlusion-aware hand-object pose estimation method based on masked autoencoders, termed as HOMAE. Specifically, we propose a target-focused masking strategy that imposes structured occlusion on regions of hand-object interaction, encouraging the model to learn context-aware features and reason about the occluded structures. We further integrate multi-scale features extracted from the decoder to predict a signed distance field (SDF), capturing both global context and fine-grained geometry. To enhance geometric perception, we combine the implicit SDF with an explicit point cloud derived from the SDF, leveraging the complementary strengths of both representations. This fusion enables more robust handling of occluded regions by combining the global context from the SDF with the precise local geometry provided by the point cloud. Extensive experiments on challenging DexYCB and HO3Dv2 benchmarks demonstrate that HOMAE achieves state-of-the-art performance in hand-object pose estimation. We will release our code and model.

cs.CV↗