arXiv subjects

Xiaobao Wang

Publications and source records attributed to Xiaobao Wang.

At least 37 records · Page 2Linked to original sources

arXiv2025

Active Multimodal Distillation for Few-shot Action Recognition

Owing to its rapid progress and broad application prospects, few-shot action recognition has attracted considerable interest. However, current methods are predominantly based on limited single-modal data, which does not fully exploit the potential of multimodal information. This paper presents a novel framework that actively identifies reliable modalities for each sample using task-specific contextual cues, thus significantly improving recognition performance. Our framework integrates an Active Sample Inference (ASI) module, which utilizes active inference to predict reliable modalities based on posterior distributions and subsequently organizes them accordingly. Unlike reinforcement learning, active inference replaces rewards with evidence-based preferences, making more stable predictions. Additionally, we introduce an active mutual distillation module that enhances the representation learning of less reliable modalities by transferring knowledge from more reliable ones. Adaptive multimodal inference is employed during the meta-test to assign higher weights to reliable modalities. Extensive experiments across multiple benchmarks demonstrate that our method significantly outperforms existing approaches.

cs.CV↗

arXiv2025

Single-Node Trigger Backdoor Attacks in Graph-Based Recommendation Systems

Graph recommendation systems have been widely studied due to their ability to effectively capture the complex interactions between users and items. However, these systems also exhibit certain vulnerabilities when faced with attacks. The prevailing shilling attack methods typically manipulate recommendation results by injecting a large number of fake nodes and edges. However, such attack strategies face two primary challenges: low stealth and high destructiveness. To address these challenges, this paper proposes a novel graph backdoor attack method that aims to enhance the exposure of target items to the target user in a covert manner, without affecting other unrelated nodes. Specifically, we design a single-node trigger generator, which can effectively expose multiple target items to the target user by inserting only one fake user node. Additionally, we introduce constraint conditions between the target nodes and irrelevant nodes to mitigate the impact of fake nodes on the recommendation system's performance. Experimental results show that the exposure of the target items reaches no less than 50% in 99% of the target users, while the impact on the recommendation system's performance is controlled within approximately 5%.

cs.AI↗

arXiv2025

Integration of Old and New Knowledge for Generalized Intent Discovery: A Consistency-driven Prototype-Prompting Framework

Intent detection aims to identify user intents from natural language inputs, where supervised methods rely heavily on labeled in-domain (IND) data and struggle with out-of-domain (OOD) intents, limiting their practical applicability. Generalized Intent Discovery (GID) addresses this by leveraging unlabeled OOD data to discover new intents without additional annotation. However, existing methods focus solely on clustering unsupervised data while neglecting domain adaptation. Therefore, we propose a consistency-driven prototype-prompting framework for GID from the perspective of integrating old and new knowledge, which includes a prototype-prompting framework for transferring old knowledge from external sources, and a hierarchical consistency constraint for learning new knowledge from target domains. We conducted extensive experiments and the results show that our method significantly outperforms all baseline methods, achieving state-of-the-art results, which strongly demonstrates the effectiveness and generalization of our methods. Our source code is publicly available at https://github.com/smileix/cpp.

cs.CL↗

arXiv2025

Rethinking Contrastive Learning in Graph Anomaly Detection: A Clean-View Perspective

Graph anomaly detection aims to identify unusual patterns in graph-based data, with wide applications in fields such as web security and financial fraud detection. Existing methods typically rely on contrastive learning, assuming that a lower similarity between a node and its local subgraph indicates abnormality. However, these approaches overlook a crucial limitation: the presence of interfering edges invalidates this assumption, since it introduces disruptive noise that compromises the contrastive learning process. Consequently, this limitation impairs the ability to effectively learn meaningful representations of normal patterns, leading to suboptimal detection performance. To address this issue, we propose a Clean-View Enhanced Graph Anomaly Detection framework (CVGAD), which includes a multi-scale anomaly awareness module to identify key sources of interference in the contrastive learning process. Moreover, to mitigate bias from the one-step edge removal process, we introduce a novel progressive purification module. This module incrementally refines the graph by iteratively identifying and removing interfering edges, thereby enhancing model performance. Extensive experiments on five benchmark datasets validate the effectiveness of our approach.

cs.LG↗

arXiv2024

Enriching Multimodal Sentiment Analysis through Textual Emotional Descriptions of Visual-Audio Content

Multimodal Sentiment Analysis (MSA) stands as a critical research frontier, seeking to comprehensively unravel human emotions by amalgamating text, audio, and visual data. Yet, discerning subtle emotional nuances within audio and video expressions poses a formidable challenge, particularly when emotional polarities across various segments appear similar. In this paper, our objective is to spotlight emotion-relevant attributes of audio and visual modalities to facilitate multimodal fusion in the context of nuanced emotional shifts in visual-audio scenarios. To this end, we introduce DEVA, a progressive fusion framework founded on textual sentiment descriptions aimed at accentuating emotional features of visual-audio content. DEVA employs an Emotional Description Generator (EDG) to transmute raw audio and visual data into textualized sentiment descriptions, thereby amplifying their emotional characteristics. These descriptions are then integrated with the source data to yield richer, enhanced features. Furthermore, DEVA incorporates the Text-guided Progressive Fusion Module (TPF), leveraging varying levels of text as a core modality guide. This module progressively fuses visual-audio minor modalities to alleviate disparities between text and visual-audio modalities. Experimental results on widely used sentiment analysis benchmark datasets, including MOSI, MOSEI, and CH-SIMS, underscore significant enhancements compared to state-of-the-art models. Moreover, fine-grained emotion experiments corroborate the robust sensitivity of DEVA to subtle emotional variations.

cs.CV↗

arXiv2024

Progressive Residual Extraction based Pre-training for Speech Representation Learning

Self-supervised learning (SSL) has garnered significant attention in speech processing, excelling in linguistic tasks such as speech recognition. However, jointly improving the performance of pre-trained models on various downstream tasks, each requiring different speech information, poses significant challenges. To this purpose, we propose a progressive residual extraction based self-supervised learning method, named ProgRE. Specifically, we introduce two lightweight and specialized task modules into an encoder-style SSL backbone to enhance its ability to extract pitch variation and speaker information from speech. Furthermore, to prevent the interference of reinforced pitch variation and speaker information with irrelevant content information learning, we residually remove the information extracted by these two modules from the main branch. The main branch is then trained using HuBERT's speech masking prediction to ensure the performance of the Transformer's deep-layer features on content tasks. In this way, we can progressively extract pitch variation, speaker, and content representations from the input speech. Finally, we can combine multiple representations with diverse speech information using different layer weights to obtain task-specific representations for various downstream tasks. Experimental results indicate that our proposed method achieves joint performance improvements on various tasks, such as speaker identification, speech recognition, emotion recognition, speech enhancement, and voice conversion, compared to excellent SSL methods such as wav2vec2.0, HuBERT, and WavLM.

eess.AS↗

arXiv2024

Is a photon ring invariably a closed structure?

In this study, we investigate the image of a rotating compact object (CO) illuminated by a geometrically thin, optically thin disk on the equatorial plane. As the radius of the CO's surface fluctuates, the CO may partially or entirely obscure the photon region. We observe that the perceived photon ring may exhibit discontinuities, deviating from a closed structure, and may even disappear entirely. We find that the disruption and disappearance of the photon ring are dependent on the observational angle$-$a novel phenomenon not previously observed in black hole imaging studies. Our study reveals that while the factors influencing this unique photon ring phenomenon are diverse and the outcomes complex, we can provide a clear and comprehensive explanation of the physical essence and variation trends of this phenomenon. We do this by introducing and analyzing the properties and interrelationships of three characteristic functions, $\tildeη$, $η_o$, and $η_s$ related to the photon impact parameters. Additionally, our analysis of the intensity cuts and inner shadows of the images uncovers patterns that differ significantly from the shadow curve.

gr-qc↗

arXiv2024

AIMDiT: Modality Augmentation and Interaction via Multimodal Dimension Transformation for Emotion Recognition in Conversations

Emotion Recognition in Conversations (ERC) is a popular task in natural language processing, which aims to recognize the emotional state of the speaker in conversations. While current research primarily emphasizes contextual modeling, there exists a dearth of investigation into effective multimodal fusion methods. We propose a novel framework called AIMDiT to solve the problem of multimodal fusion of deep features. Specifically, we design a Modality Augmentation Network which performs rich representation learning through dimension transformation of different modalities and parameter-efficient inception block. On the other hand, the Modality Interaction Network performs interaction fusion of extracted inter-modal features and intra-modal features. Experiments conducted using our AIMDiT framework on the public benchmark dataset MELD reveal 2.34% and 2.87% improvements in terms of the Acc-7 and w-F1 metrics compared to the state-of-the-art (SOTA) models.

cs.MM↗

arXiv2023

speech and noise dual-stream spectrogram refine network with speech distortion loss for robust speech recognition

In recent years, the joint training of speech enhancement front-end and automatic speech recognition (ASR) back-end has been widely used to improve the robustness of ASR systems. Traditional joint training methods only use enhanced speech as input for the backend. However, it is difficult for speech enhancement systems to directly separate speech from input due to the diverse types of noise with different intensities. Furthermore, speech distortion and residual noise are often observed in enhanced speech, and the distortion of speech and noise is different. Most existing methods focus on fusing enhanced and noisy features to address this issue. In this paper, we propose a dual-stream spectrogram refine network to simultaneously refine the speech and noise and decouple the noise from the noisy input. Our proposed method can achieve better performance with a relative 8.6% CER reduction.

cs.SD↗

arXiv2023

Traversable Thin-shell Wormhole in the 4D Einstein-Gauss-Bonnet Theory

This work investigates the spherically symmetric thin-shell wormhole solutions in four-dimensional Einstein-Gauss-Bonnet theory and explores their stabilities under radial, linear perturbations. These solutions are typically traversable and characterized by a thin-shell throat in accordance with Israel's junction conditions. In asymptotically flat and AdS spacetimes with a negative Gauss-Bonnet coupling constant, stable neutral wormholes are encountered when the magnitude of the coupling constant becomes significant. The throats of such wormholes are sustained by ordinary matter and possess finite radii. In asymptotically dS spacetimes, no stable neutral wormhole featuring ordinary matter is observed. On the other hand, for positive Gauss-Bonnet coupling constant, stable thin-shell wormhole solutions can be established when the throats are exclusively supported by exotic matter. Moreover, stable charged wormholes comprised of ordinary matter are found universally in the asymptotically flat, AdS, and dS spacetimes. Unlike their neutral counterparts, the throat radii of such charged wormholes can be arbitrarily small. However, as the charge becomes more significant, such solutions only remain stable when supported by exotic matter.

gr-qc↗

arXiv2022

Measuring neutron skin by grazing isobaric collisions

Neutron skin thickness ($Δr_{\rm np}$) of nuclei and the inferred nuclear symmetry energy are of critical importance to nuclear physics and astrophysics. It is traditionally measured by nuclear processes with significant theoretical uncertainties. We recently proposed an indirect measurement of the $Δr_{\rm np}$ by charged hadron multiplicities in central isobaric collisions at relativistic energies, which are sensitive to nuclear densities. In this Letter, we propose a direct measurement of the $Δr_{\rm np}$ by using net-charge multiplicities in ultra-peripheral (grazing) collisions of those isobars, under the assumption that they are simple superimposition of nucleon-nucleon interactions. We illustrate this novel approach by the TRENTO and URQMD models.

nucl-th↗

arXiv2021

Determine the neutron skin type by relativistic isobaric collisions

The effects of neutron skin on the multiplicity ($N_{\rm ch}$) and eccentricity($ε_2$) in relativistic $^{96}_{44}$Ru+$^{96}_{44}$Ru and $^{96}_{40}$Zr+$^{96}_{40}$Zr collisions at $\sqrt{s_{_{\rm NN}}}=200$ GeV are investigated with the Trento model. It is found that the Ru+Ru/Zr+Zr ratios of the $N_{\rm ch}$ distributions and $ε_{2}$ in mid-central collisions are exquisitely sensitive to the neutron skin type (skin vs.~halo). The state-of-the-art calculations by energy density functional theory (DFT) favor the halo-type neutron skin and can soon be confronted by experimental data. It is demonstrated that the halo-type density can serve as a good surrogate for the DFT density, and thus can be efficiently employed to probe nuclear deformities by using elliptic flow data in central collisions. We provide hereby a proof-of-principle venue to simultaneously determine the neutron skin type, thickness, and nuclear deformity.

nucl-th↗

arXiv2020

Probing the neutron skin with ultrarelativistic isobaric collisions

Particle production in ultrarelativistic heavy ion collisions depends on the details of the nucleon density distributions in the colliding nuclei. We demonstrate that the charged hadron multiplicity distributions in isobaric collisions at ultrarelativistic energies provide a novel approach to determine the poorly known neutron density distributions and thus the neutron skin thickness in finite nuclei, which can in turn put stringent constraints on the nuclear symmetry energy.

nucl-th↗

arXiv2020

Novel shadows from the asymmetric thin-shell wormhole

For dark compact objects such as black holes or wormholes, the shadow size has long been thought to be determined by the unstable photon sphere (region). However, by considering the asymmetric thin-shell wormhole (ATSW) model, we find that the impact parameter of the null geodesics is discontinuous through the wormhole in general and hence we identify novel shadows whose sizes are dependent of the photon sphere in the other side of the spacetime. The novel shadows appear in three cases: (A2) The observer's spacetime contains a photon sphere and the mass parameter is smaller than that of the opposite side; (B1, B2) there' s no photon sphere no matter which mass parameter is bigger. In particular, comparing with the black hole, the wormhole shadow size is always smaller and their difference is significant in most cases, which provides a potential way to observe wormholes directly through Event Horizon Telescope with better detection capability in the future.

gr-qc↗

arXiv2019

Multi-phase transport model predictions of isobaric collisions with nuclear structures from density functional theory

Isobaric $^{96}_{44}$Ru+$^{96}_{44}$Ru and $^{96}_{40}$Zr+$^{96}_{40}$Zr collisions were performed at the Relativistic Heavy Ion Collider in 2018. Using the "a multi-phase transport" model with nuclear structures calculated by the density functional theory (DFT), we make predictions for the charged hadron multiplicity distributions and elliptic azimuthal anisotropies in these collisions. Emphases are put on the relative differences between the two collision systems that can decisively discriminate DFT nuclear distributions from the commonly used Woods-Saxon densities.

nucl-th↗

arXiv2019

Critical phenomena in gravitational collapse of Husain-Martinez-Nunez scalar field

We construct analytical models to study the critical phenomena in gravitational collapse of the Husain-Martinez-Nunez massless scalar field. We first use the cut-and-paste technique to match the conformally flat solution ($c=0$ ) onto an outgoing Vaidya solution. To guarantee the continuity of the metric and the extrinsic curvature, we prove that the two solutions must be joined at a null hypersurface and the metric function in Vaidya spacetime must satisfy some constraints. We find that the mass of the black hole in the resulting spacetime takes the form $M\propto (p-p^*)^γ$, where the critical exponent $γ$ is equal to $0.5$. For the case $c\neq 0$, we show that the scalar field must be joined onto two pieces of Vaidya spacetimes to avoid a naked singularity. We also derive the power-law mass formula with $γ=0.5$. Compared with previous analytical models constructed from a different scalar field with continuous self-similarity, we obtain the same value of $γ$. However, we show that the solution with $c\neq 0$ is not self-similar. Therefore, we provide a rare example that a scalar field without self-similarity also possesses the features of critical collapse.

gr-qc↗

arXiv2018

Importance of isobar density distributions on the chiral magnetic effect search

Under the approximate chiral symmetry restoration, quark interactions with topological gluon fields in quantum chromodynamics can induce chirality imbalance and parity violation in local domains. An electric charge separation ({\sc cs}) could be generated along the direction of a strong magnetic field ({\bf B}), a phenomenon called the chiral magnetic effect ({\sc cme}). {\sc cs} measurements by azimuthal correlators are contaminated by a major background from elliptic flow anisotropy ($v_2$). Isobaric $^{96}_{44}$Ru+$^{96}_{44}$Ru and $^{96}_{40}$Zr+$^{96}_{40}$Zr collisions have been proposed to identify the {\sc cme} (expected to differ between the two systems) out of the background (expected to be almost the same). We show, by using the density-functional calculated proton and neutron distributions, that these expectations may not hold as originally anticipated, because the two systems may have sizable differences in eccentricity and $v_2$ and because their difference in {\bf B} may suffer from large uncertainties.

nucl-th↗

arXiv2018

Re-examining the premise of isobaric collisions and a novel method to measure the chiral magnetic effect

In this proceeding we will show that the expectations of the isobaric $^{96}_{44}\mathrm{Ru}+^{96}_{44}\mathrm{Ru}$ and $^{96}_{40}\mathrm{Zr}+^{96}_{40}\mathrm{Zr}$ collisions on chiral magnetic effect (CME) search may not hold as originally anticipated due to large uncertainties in the isobaric nuclear structures. We demonstrate this using Woods-Saxon densities and the proton and neutron densities calculated by the density functional theory. Furthermore, a novel method is proposed to gauge background and possible CME contributions in the same system, intrinsically better than the isobaric collisions of two different systems. We illustrate the method with Monte Carlo Glauber and AMPT (A Multi-Phase Transport) simulations.

nucl-th↗