arXiv subjects

Dong Wang

Publications and source records attributed to Dong Wang.

At least 55 records · Page 3Linked to original sources

arXiv2026

On the Interaction Between Model Compression and Test-Time Adaptation

Deep neural networks deployed in the wild must be both efficient and adaptable, requiring model compression and test-time adaptation (TTA). While both are well studied in isolation, their interaction remains poorly understood. We systematically analyze how structured compression affects a model's ability to adapt under distribution shift. Using ResNet-18 and ViT-Base on CIFAR-10-C and ImageNet-C, we evaluate multiple compression methods combined with standard TTA techniques. We introduce a diagnostic framework that examines representational expressivity and adaptation subspace compatibility. Our results reveal a consistent gap: although compressed models retain high accuracy under supervised adaptation, their TTA performance degrades significantly with increasing compression. We show that this stems from reduced representational diversity and structural constraints that limit recoverability. These effects strongly depend on the compression method, highlighting the need to design compression strategies that preserve adaptability.

cs.LG↗

arXiv2026

The advantages of extended nonreciprocal quantum batteries

This study investigates the performance of extended nonreciprocal quantum batteries (QBs), as well as its advantages in energy storage and energy transfer compared to reciprocal charging and the original nonreciprocal batteries. After analyzing the detuning between the charging system and the external pump, we discover that resonance is a key factor in maintaining high-energy batteries and high charging power; furthermore, the detuning of the charger or battery determines the stability of the charging process for different structures. Research on steady-state energy storage in batteries revealed that single-threaded or multi-threaded charging can achieve nearly infinite energy storage in weakly localized environments, thereby demonstrating the significant energy advantages of extended nonreciprocal quantum batteries. Finally, by considering the energy distribution within the charging system, we observe that nonreciprocal charging offers energy transfer advantages unmatched by reciprocal charging; the former achieves a comprehensive balance between charging cost and energy storage capacity that the latter cannot match. As a novel and superior charging protocol, our findings are expected to provide a potent reference for the promotion and practical implementation of nonreciprocal charging.

quant-ph↗

arXiv2026

PropUQ-MAS: Propagation-Aware Uncertainty Quantification for LLM Multi-Agent Systems

LLM-based multi-agent systems (MAS) solve complex tasks through communication among role-specialized agents. However, inter-agent dependencies introduce reliability risks beyond isolated agent failures. For instance, errors in intermediate messages could be inherited and amplified by downstream agents. Existing uncertainty quantification (UQ) methods mainly target isolated responses or single-agent reasoning, and therefore fail to capture uncertainty propagation in MAS. To this end, we propose PropUQ-MAS, an error propagation-aware UQ framework that represents MAS execution as a communication-structured graph and estimates each step's reliability by combining local uncertainty with uncertainty inherited from upstream messages. Extensive experiments demonstrate that PropUQ-MAS consistently improves UQ in MAS, with average relative gains of +6.10% in AUROC and +47.58% in PRR.

cs.MA↗

arXiv2026

PhyGile: Physics-Prefix Guided Motion Generation for Agile General Humanoid Motion Tracking

Humanoid robots are expected to execute agile and expressive whole-body motions in real-world settings. Existing text-to-motion generation models are predominantly trained on captured human motion datasets, whose priors assume human biomechanics, actuation, mass distribution, and contact strategies. When such motions are directly retargeted to humanoid robots, the resulting trajectories may satisfy geometric constraints (e.g., joint limits and pose continuity) and appear kinematically reasonable. However, they frequently violate the physical feasibility required for real-world execution. To address these issues, we present PhyGile, a unified framework that closes the loop between robot-native motion generation and General Motion Tracking (GMT). PhyGile performs physics-prefix-guided robot-native motion generation at inference time, directly generating robot-native motions in a 262-dimensional skeletal space with physics-guided prefixes, thereby eliminating inference-time retargeting artifacts and reducing generation-execution discrepancies. Before physics-prefix adaptation, we train the GMT controller with a curriculum-based mixture-of-experts scheme, followed by post-training on unlabeled motion data to improve robustness over large-scale robot motions. During physics-prefix adaptation, the GMT controller is further fine-tuned with generated objectives under physics-derived prefixes, enabling agile and stable execution of complex motions on real robots. Extensive offline and real-robot experiments demonstrate that PhyGile expands the frontier of text-driven humanoid control, enabling stable tracking of agile, highly difficult whole-body motions that go well beyond walking and low-dynamic motions typically achieved by prior methods.

cs.RO↗

arXiv2026

Beyond Truth Discovery: A Two-Stage Framework to Assess the Severity of False Claim during Disasters

False information spreads rapidly on social media during disasters and can undermine emergency response efforts, public trust, and crisis communication. Existing research primarily focuses on determining whether social media posts contain false information, but provides limited insight into the specific false claims embedded within posts and the severity of individual false claims. To address the limitations, we propose a two-stage framework to assess the severity of false claims during disasters. In the first stage, we develop a false claim extraction agent that identifies false claims from multimodal social media posts containing text, images, videos, and links. A subsequent verification step validates extracted claims with supporting evidence. In the second stage, we define false claim severity as the combination of two complementary dimensions: believability, which determines the likelihood that a claim will be believed, and harmfulness, which captures the potential consequences if it is believed. Human annotators assess both dimensions to construct a claim-level severity benchmark using false claims extracted from Reddit posts related to hurricanes and wildfires. Building upon this benchmark, we investigate false claim severity assessment as a human-AI alignment problem, evaluating whether models can reproduce human judgments under a shared evaluation rubric rather than merely predicting severity labels. Experiments on the benchmark show that traditional supervised models exhibit limited alignment with human judgments, whereas Large Language Models (LLMs) achieve substantially stronger performance. Among the evaluated strategies, in-context learning consistently achieves the strongest alignment with human judgments, highlighting the importance of human examples and shared decision criteria for severity assessment.

cs.SI↗

arXiv2026

HUG-VIS: A Multimodal Benchmark for Human-centered Understanding and Generation in Visual Intelligence

Visual intelligence seeks to perceive, interpret, and synthesize the visual world and is central to modern computer vision. Human-centered visual intelligence is especially demanding because it studies people as expressive, socially situated subjects whose meaning is rarely conveyed by appearance alone. It couples vision with audio and language across four representative tasks: human emotion recognition, human video generation, human voice cloning, and human video matting. Yet existing resources remain task-specific, providing modalities and annotations for individual problems rather than a shared foundation coordinating understanding and generation. This limits multimodal signal use and broader research. We address this gap with HUG-VIS, a unified benchmark for Human-centered Understanding and Generation in Visual Intelligence. It contains 8,400 seated half-body videos of 30 professional actors, each performing the same 280 emotion-action-prompt assignments under a controlled Mandarin studio protocol, with synchronized video, audio, text, and alpha mattes. We evaluate diverse open- and closed-source models across the four tasks under a unified zero-shot protocol using automatic metrics, criterion-specific mean opinion scores, and multiple cross-task analyses. Results show that (i) linguistic content dominates current emotion recognition, while purely visual affect recognition is weakest; (ii) in video generation and voice cloning, automatic metrics and human judgment agree overall but differ in their top rankings, requiring joint reporting; (iii) boundary fidelity under motion is the main remaining obstacle for human matting; and (iv) task difficulty varies across emotions, models, and metrics, with notable cross-task correlations. The dataset and results are available at https://github.com/GML-MMGroup/HUG-VIS.

cs.CV↗

arXiv2026

Quantum coherence and entanglement in wireless quantum batteries

We investigate the charging dynamics and thermodynamic performance of a wireless quantum battery system mediated by a common structured bosonic environment. By employing a unified resource-theoretic analysis, we elucidate the distinct roles of non-Markovian memory effects and coupling symmetry in regulating energy transfer. In the Markovian weak-coupling regime, we identify a transformative mechanism where the dynamic reconstruction of $l_1$-norm coherence compensates for the monotonic decay of first-order coherence to sustain energy transport. Conversely, the non-Markovian strong-coupling regime facilitates a cooperative resonance, characterized by the synchronized oscillation of entanglement and stored energy induced by environmental backflow. Furthermore, we reveal that coupling symmetry acts as a critical control parameter: while asymmetric coupling favoring the battery optimizes energy gain in memoryless environments, symmetric coupling under strong interactions unlocks a dark-state protection mechanism, effectively trapping energy within a decoherence-free subspace. Finally, a thermodynamic analysis based on ergotropy demonstrates that first-order coherence establishes the activation threshold for incoherent work, whereas $l_1$-norm coherence serves as the explicit fuel for coherent work extraction. These findings provide a refined theoretical framework for engineering environment-assisted quantum energy storage devices.

quant-ph↗

arXiv2026

SCoRD: Semantic-Assisted Continual Retriever-Reranker Distillation for LLM-Based Recommendation

Recommendation systems increasingly adopt a two-stage pipeline, where an ID-based retriever retrieves candidates and an LLM-based reranker refines their rankings. To improve retrieval quality, reranker-to-retriever distillation is commonly used to transfer the reranker's knowledge to the retriever. For practical deployment, however, this pipeline must continually adapt to evolving interests and incoming interactions. A naive solution is to repeatedly update the LLM reranker and distill its latest knowledge, but this incurs prohibitive costs. Updating the retriever alone is cheaper, but its limited capacity makes adaptation from sparse data difficult. We propose SCoRD, a continual knowledge distillation framework for LLM-based reranking pipelines under a non-stationary data stream. SCoRD introduces a semantic reasoning assistant that distills the LLM's ability to infer underlying user intents into reusable intent-level guidance. It selectively distills reranker knowledge to the retriever on low-confidence sequences, guides retriever-only updates without repeated LLM inference, and feeds retriever-derived representations and intent-drift signals back to the reranker. Experiments on real-world datasets show that SCoRD enables effective and efficient retriever-reranker co-adaptation.

cs.IR↗

arXiv2026

Dynamical redistribution of quantum resources in tree-level Bhabha scattering

The fundamental interactions governed by quantum electrodynamics (QED) are intrinsically rich in quantum resources, yet how these resources dynamically redistribute during relativistic scattering is still not fully understood. In this work, we systematically investigate the tree-level Bhabha scattering process (e^- e^+ \rightarrow e^- e^+) within the framework of quantum resource theory, revealing how QED kinematics and Feynman amplitudes strictly dictate resource redistribution. Specifically, we demonstrate a strict anti-correlation between entropic uncertainty and dynamically generated entanglement across diverse initial states. We find that mass-induced single-helicity-flip transitions cause a pronounced geometric symmetry breaking in the non-relativistic regime, whereas the restoration of chiral symmetry in the ultra-relativistic limit ensures strict symmetry about the backward scattering angle. Furthermore, we analytically establish a rigorous equivalence between local wave-particle duality and global bipartite quantum coherence. Finally, evaluating the trade-off between local duality and Bell nonlocality, we show that in the ultra-relativistic limit, transverse scattering of basic factorized states equalizes the s- and t-channel amplitudes to optimize non-local correlations. However, pre-existing local coherence inevitably disrupts this delicate kinematic balance, significantly suppressing the Bell parameter and preventing the maximal violation of local realism. Therefore, we believe the present results provide deeper understanding of the fundamental quantum nature of QED processes.

quant-ph↗

arXiv2026

A weighted quantile filter based framework for interface optimal design problems

We present a robust and efficient numerical framework based on a median filter scheme for solving a broad class of interface optimal design problems, from image segmentation to topology optimization. A key innovation of our work is the extension of the binary scheme into a level-set scheme via a weighted quantile interpretation. Unlike traditional binary iterative convolution-thresholding method (ICTM), this continuous weighted quantile filter scheme effectively overcomes the pinning effect caused by spatial discretization, achieving interface evolution even with small time steps. We also provide a rigorous theoretical analysis, proving the unconditional energy stability of the iterative scheme. Furthermore, we prove that for a wide class of data fidelity terms, the convex relaxation inherently enforces a binary solution, justifying the effectiveness of the method without explicit penalization. Numerical experiments on the Chan--Vese model, the local intensity fitting (LIF) model, and topology optimization in Stokes flow demonstrate that the proposed efficient continuous framework effectively eliminates the pinning effect, guarantees unconditional energy stability, and accurately converges to binary solutions.

math.NA↗

arXiv2026

Structured generalized sliced Wasserstein distance for keV X-ray polarization analysis with Gas Pixel Detector

Because of the special angular distribution of excited electrons by the photoelectric effect, the Gas Pixel Detector (GPD) is effective in measuring keV X-ray polarization of astrophysical events (e.g. gamma-ray bursts), by capturing ionization tracks of excited electrons as polarized images. Traditionally, the emission angles of photoelectrons are extracted from polarized images first, and statistics are then performed on these angles to infer the polarization direction and intensity. However, observation with the wide field of view requires the incident angle of X-rays not directly attainable through the traditional analysis process. In this paper, we propose using the generalized sliced Wasserstein (GSW) distance, projected by neural networks with random weights, as a completely data-driven approach to analyze X-ray polarization based on two-dimensional polarized images. We find the structures of the randomized neural networks matter when focusing on different aspects of the polarized images, and take advantage of the discrimination abilities by different neural network structures. The proposed method, named the structured GSW distance, successfully distinguishes polarized images with different configurations of incident angles and polarization directions. Furthermore, we build a simplified statistical model based on the von Mises distribution and the circular Wasserstein distance and compare the model against the proposed method, showing their high consistency. The computational method reported in this paper may benefit GPD-based polarimetry in astroparticle experiments and also pattern analysis on raw data from pixel detectors.

physics.data-an↗

arXiv2026

Eigenvector distribution of random matrices under critical finite-rank deformations

We investigate the eigenvector distribution at the soft edge for Gaussian random matrices with finite-rank deformations, in the critical regime of BBP transition. For finite-rank deformations of the GOE and GUE with critical spikes, we find that the squared overlap between a leading eigenvector and a spike, rescaled by $N^{1/3}$, converges weakly to the negative reciprocal of the derivative of an Airy-Green function evaluated at the corresponding soft-edge root. For the rank-one critically spiked Gaussian $β$-ensemble, $β>0$, we obtain an analogous result involving an Airy-Green function. In both cases, the Airy-Green functions are generalizations of the one introduced by Bykhovskaya--Gorin--Sodin \cite{Bykhovskaya-Gorin-Sodin25}. The proofs are both based on an eigenvector--eigenvalue identity and a resolvent-differentiation mechanism.

math.PR↗

arXiv2026

Uncertainty-Aware Variational Reward Factorization via Probabilistic Preference Bases for LLM Personalization

Reward factorization personalizes large language models (LLMs) by decomposing rewards into shared basis functions and user-specific weights. Yet, existing methods estimate user weights from scarce data in isolation and as deterministic points, leading to inaccurate and unreliable inference. We introduce Variational Reward Factorization (VRF), an uncertainty-aware framework that represents each user's preferences as a variational distribution in a shared preference space. VRF infers user distributions via a variational encoder, derives weights through Wasserstein distance matching with shared probabilistic bases, and downweights uncertain estimates through a variance-attenuated loss. On three benchmarks, VRF outperforms all baselines across seen and unseen users, few-shot scenarios, and varying uncertainty levels, with gains extending to downstream alignment. Our code is available at https://github.com/Gyu-Seok-Lee/VRF_COLM26.

cs.CL↗

arXiv2026

AnyTrack: Unifying Visual Object Tracking with Any Modalities

Visual object tracking aims to continuously locate specific targets within sequential frames, evolving from single-modal methods to multi-modal ones. However, existing multi-modal trackers are typically designed for fixed modality combinations, requiring separate models for different inputs. This leads to a poor adaptability to missing or imperfect modalities, and limited generalization. To address these issues, we propose a novel unified framework called AnyTrack for object tracking with any modalities. Specifically, we design a Modality-aware Interaction Module (MIM) to facilitate dynamic interaction across diverse modalities. This module bridges modality discrepancies and aggregates temporal cues to maintain spatio-temporal consistency during cross-modal interaction. Furthermore, we introduce a Context Understanding Module (CUM) to establish spatial correspondence between visual features and target locations via global-local prompts. This module employs target-aware context modeling to enhance foreground-background discrimination for precise localization. Finally, to support the training and evaluation under diverse modalities, we extend existing multi-modal object tracking benchmarks by incorporating grayscale images, language descriptions, and audio clips. Extensive experiments with both complete and missing modality settings demonstrate that our AnyTrack achieves state-of-the-art performance, validating its effectiveness and flexibility. The source code is available at https://github.com/IdolLab/AnyTrack.

cs.CV↗

arXiv2026

Agent-Based Test Assertion Generation via Diverse Perspective Aggregation

Test assertions are critical elements of unit tests, serving as checkpoints to validate expected behavior and ensure software correctness. Numerous techniques have been proposed to automate assertion generation, with recent progress notably driven by large language models (LLMs). Despite the promise, existing approaches such as ChatAssert suffer from modest accuracy, heavy reliance on oversampling, and vulnerability to model randomness due to one-shot prompting. To address these limitations, we propose AssertMate, a novel agent-based assertion generation framework that enhances the quality and reliability of LLM-generated assertions through three key components: (1) actual value construction that identifies assertion targets via static analysis and type-aware heuristics; (2) multi-perspective expected value prediction using code generation, retrieval-augmented generation (RAG), and chain-of-thought (CoT) reasoning agents; and (3) an LLM-as-a-Judge collaboration mechanism to select the most appropriate assertion. Evaluation on the Defects4J benchmark demonstrates that AssertMate significantly outperforms state-of-the-art techniques in compilation success and pass rates, along with substantially higher bug detection capabilities. Integration with EvoSuite further validates AssertMate's practicality, yielding superior mutation coverage and kill counts. Ablation studies reveal that each of the three components makes a significant and complementary contribution to the overall performance. This work affirms the great potential of aggregating diverse perspectives to enhance the effectiveness of LLM-based assertion generation.

cs.SE↗

arXiv2026

MeSS: City Mesh-Guided Outdoor Scene Generation with Cross-View Consistent Diffusion

Mesh models have become increasingly accessible for numerous cities; however, the lack of realistic textures restricts their application in virtual urban navigation and autonomous driving. To address this, this paper proposes MeSS (Meshbased Scene Synthesis) for generating high-quality, styleconsistent outdoor scenes with city mesh models serving as the geometric prior. While image and video diffusion models can leverage spatial layouts (such as depth maps or HD maps) as control conditions to generate street-level perspective views, they are not directly applicable to 3D scene generation. Video diffusion models excel at synthesizing consistent view sequences that depict scenes but often struggle to adhere to predefined camera paths or align accurately with rendered control videos. In contrast, image diffusion models, though unable to guarantee cross-view visual consistency, can produce more geometry-aligned results when combined with ControlNet. Building on this insight, our approach enhances image diffusion models by improving cross-view consistency. The pipeline comprises three key stages: first, we generate geometrically consistent sparse views using Cascaded Outpainting ControlNets; second, we propagate denser intermediate views via a component dubbed AGInpaint; and third, we globally eliminate visual inconsistencies (e.g., varying exposure) using the GCAlign module. Concurrently with generation, a 3D Gaussian Splatting (3DGS) scene is reconstructed by initializing Gaussian balls on the mesh surface. Our method outperforms existing approaches in both geometric alignment and generation quality. Once synthesized, the scene can be rendered in diverse styles through relighting and style transfer techniques. project page: https://albertchen98.github.io/mess/

cs.CV↗

arXiv2026

Reinforcement-Learned Electric-Field Sensing with Asymmetrically Blockaded Rydberg Arrays

We present a reinforcement learning-optimized Rydberg electrometer based on the asymmetric blockade effect and achieve high-sensitivity electric field sensing in Rydberg arrays. Microwave dressing induces asymmetric blockade to suppress interactions between target atoms, while keeping the coupling between the central control atom and target atoms field-tunable near Förster resonance. The field-regulated blockade radius affects the detectable atomic population signals, thereby enabling electric field sensing via state-selective readout. In planar atomic arrays, classical Fisher information exhibits near-quadratic scaling with atom number and approaches the Heisenberg limit. Reinforcement learning-designed composite pulses greatly enhance quantum Fisher information by up to one order of magnitude compared with single $π$ pulses. We further establish a compact six-atom spherical configuration for vector electrometry, in which field orientation is extracted from calibrated axial populations, and weak bias fields eliminate dipole-dipole-induced sign and magic-angle ambiguities. Numerical tests against Rabi frequency deviation, positional error, residual inter-target coupling and projection noise demonstrate the reliability of this scheme. This work provides an experimentally viable approach to realize high-precision three-dimensional Rydberg electric field sensing.

quant-ph↗

arXiv2026

Transducer leakage error suppression using invariant-based shortcut

We present a method for suppressing transducer leakage errors in spin-superconducting hybrid quantum systems with the theory of optimal invariant-based shortcut. By mediated virtual photons as a transducer to exchange the energy between the spin qubit and a transmon qubit, the fidelity of the population of the final state features a broad range above 99\% under the influence of leakage error. The leakage probability from computational subspace to non-computational subspace can be effectively suppressed at a lowest value with $0.01$. Based on the optimized pulse control designed by the invariant-based inverse engineering, the high-fidelity quantum iSWAP gate operations and entanglement state preparation within the computational subspace are achieved. {Compared to the traditional $π$ pulse, derivative removal by adiabatic gate, counter-diabatic shortcut schemes, and limited-memory Broyden-Fletcher-Goldfarb-Shanno gradient ascent pulse engineering, the optimized shortcut scheme can still achieve a high fidelity with 99\% in the presence of decoherence and control error.} When taking into account the possibility of leakage errors in actual situations, our solution can still largely resist the influence of control errors. The results provides a feasible path for precisely manipulating the quantum state of hybrid quantum systems.

quant-ph↗