Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Suggested from co-authorship
- Qiang JiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works167 from public data
- Invisible Backdoor Attack with Sample-Specific Triggers721
Inspired by the recent advance in DNN-based image steganography, sample-specific invisible additive noises as backdoor triggers are generated by encoding an attacker-specified string into benign images through an encoder-decoder network.
- Bi-Real Net: Enhancing the Performance of 1-Bit CNNs with Improved Representational Capability and Advanced Training Algorithm662
A novel model, dubbed Bi-Real net, which connects the real activations (after the 1-bit convolution and/or BatchNorm layer, before the sign function) to activations of the consecutive block, through an identity shortcut is proposed, which achieves up to 10% higher top-1 accuracy with more memory saving and lower computational cost.
- Learning to Compose Dynamic Tree Structures for Visual Contexts623
A hybrid learning procedure is developed which integrates end-task supervised learning and the tree structure reinforcement learning, where the former's evaluation result serves as a self-critic for the latter's structure exploration.
- 496
- 479
- Efficient Decision-Based Black-Box Adversarial Attacks on Face Recognition478
This paper evaluates the robustness of state-of-the-art face recognition models in the decision-based black-box attack setting, where the attackers have no access to the model parameters and gradients, but can only acquire hard-label predictions by sending queries to the target model.
- Target-Aware Deep Tracking445
This paper develops a regression loss and a ranking loss to guide the generation of target-active and scale-sensitive features and proposes a novel scheme to learn target-aware features, which can better recognize the targets undergoing significant appearance variations than pre-trained deep features.
- UCF: Uncovering Common Features for Generalizable Deepfake Detection341
A disentanglement framework that decomposes image information into three distinct components: forgery-irrelevant, method-specific forgery, and common forgery features is proposed and can perform superior generalization than current state-of-the-art methods.
- DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection309
This work presents the first comprehensive benchmark for deepfake detection, called DeepfakeBench, which offers three key contributions: a unified data management system to ensure consistent input across all detectors, an integrated framework for state-of-the-art methods implementation, and standardized evaluation metrics and protocols to promote transparency and reproducibility.
- Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection237
This work proposes a simple yet effective detector called LSDA (Latent Space Data Augmentation), based on a heuristic idea: representations with a wider variety of forgeries should be able to learn a more generalizable decision boundary, thereby mitigating the overfitting of method-specific features.
- BackdoorBench: A Comprehensive Benchmark of Backdoor Learning237
A comprehensive benchmark of backdoor learning called BackdoorBench, which consists of an extensible modular-based codebase and a standardized protocol of complete backdoor learning, and provides comprehensive evaluations of every pair of 8 attacks against 9 defenses, with 5 poisoning ratios, thus 8,000 pairs of evaluations in total.
- CNN in MRF: Video Object Segmentation via Inference in a CNN-Based Higher-Order Spatio-Temporal MRF218
A novel CNN-embedded algorithm to perform approximate inference in the MRF, which outperforms the winning entries of the DAVIS 2017 Challenge, without resorting to model ensembling or any dedicated detectors.
- LAS-AT: Adversarial Training with Learnable Attack Strategy196
A novel framework for adversarial training is proposed by introducing the concept of “learnable attack strategy”, dubbed LAS-AT, which learns to automatically produce attack strategies to improve the model robustness.
- 181
- SIDA: Social Media Image Deepfake Detection, Localization and Explanation with Large Multimodal Model176
This paper proposes a new image deepfake detection, localization, and explanation framework, named SIDA (Social media Image Detection, localization, and explanation Assistant), which not only discerns the authenticity of images, but also delineates tampered regions through mask prediction and provides textual explanations of the model’s judgment criteria.
- 165
- BadCLIP: Dual-Embedding Guided Backdoor Attack on Multimodal Contrastive Learning145
The BadCLIP attack is introduced, which is resistant to backdoor detection and model fine-tuning defenses and draws motivations from the perspective of the Bayesian rule and proposes a dual-embedding guided framework for backdoor attacks.
- 137
- Boosting the Transferability of Adversarial Attacks with Reverse Adversarial Perturbation136
This work proposes a novel attack method, dubbed reverse adversarial perturbation (RAP), which can find more stable adversarial examples which are less sensitive to the changes of decision boundary, mitigating the overfitting of the surrogate model.
- Backdoor Attack Against Speaker Verification132
This paper demonstrates that it is possible to inject the hidden backdoor for infecting speaker verification models by poisoning the training data and designs a clustering-based attack scheme where poisoned samples from different clusters will contain different triggers, based on the understanding of verification tasks.
- p-Box ADMM: A Versatile Framework for Integer Programming125
This paper revisits the integer programming (IP) problem, and proposes a novel and versatile framework called Alternating Direction Method of Multipliers (ADMM), which significantly outperforms existing generic IP solvers both in runtime and objective.
- Multi-label learning with missing labels for image annotation and facial action unit recognition123
This work proposes a method for multi-label learning that explicitly handles missing labels and presents an efficient algorithm to learn a parametric classifier that is applied to image annotation and facial action recognition.
- Enhancing Fine-Tuning based Backdoor Defense with Sharpness-Aware Minimization122
The FT-SAM is proposed, a novel backdoor defense paradigm that aims to shrink the norms of backdoor-related neurons by incorporating sharpness-aware minimization with fine-tuning, and provides a promising avenue for improving the robustness of machine learning models against backdoor attacks.
- Can ChatGPT Detect DeepFakes? A Study of Using Multimodal Large Language Models for Media Forensics121
The capabilities of multimodal large language models (LLMs) in DeepFake detection are investigated and it is shown that they can expose AI-generated images through careful experimental design and prompt engineering.
- 121
- Bi-Real Net: Binarizing Deep Network Towards Real-Network Performance115
Bi-Real net with a novel training algorithm is proposed with a tighter approximation to the derivative of the sign function, a magnitude-aware binarization for weight updating, a better initialization method, and a two-step scheme for training a deep network.
- 111
- 110
- 109
- Parallel Rectangle Flip Attack: A Query-based Black-box Attack against Object Detection108
Extensive experiments demonstrate that the proposed Parallel Rectangle Flip Attack (PRFA) can effectively and efficiently attack various popular object detectors, including anchor-based and anchor- free, and generate transferable adversarial examples.
- Sparse Adversarial Attack via Perturbation Factorization101
An e-cient algorithm is developed by equivalently reformulating the MIP problem as a continuous optimization problem to jointly optimize the binary selection factors and continuous perturbation magnitudes of all pixels, with a cardinality constraint on selection factors to explicitly control the degree of sparsity.
- 97
- Tencent ML-Images: A Large-Scale Multi-Label Image Database for Visual Representation Learning96
A large-scale multi-label image database with 18M images and 11K categories is built, dubbed Tencent ML-Images, to enhance the quality of visual representation of the trained CNN model and to promote other vision tasks in the research and industry community.
- Multi-label Learning with Missing Labels Using Mixed Dependency Graphs95
This work proposes two convex transductive formulations of the MLML problem, denoted as MLMG-CO and ML MG-SL respectively, and proposes two important applications, including image annotation and tag based image retrieval, which can be jointly handled using these proposed methods.
- Boosting Decision-Based Black-Box Adversarial Attacks with Random Sign Flip82
It is shown that just randomlyipping the signs of a small number of entries in adversarial perturbations can significantly boost the attack performance.
- Boosting Fast Adversarial Training With Learnable Adversarial Initialization79
Experimental evaluations on four benchmark databases demonstrate the superiority of the proposed method over state-of-the-art fast AT methods, as well as comparable robustness to advanced multi-step AT methods.
- Probabilistic Modeling of Semantic Ambiguity for Scene Graph Generation77
This work proposes a novel plug-and-play Probabilistic Uncertainty Modeling (PUM) module, which models each union region as a Gaussian distribution, whose variance measures the uncertainty of the corresponding visual content.
- Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety76
This survey provides a systematic review of current safety research on large models, covering Vision Foundation Models, Large Language Models, Vision-Language Pre-training models, Vision-Language Models, Diffusion Models, and large-model-powered Agents, and presents a comprehensive taxonomy of safety threats to these models.
- 74
- Shared Adversarial Unlearning: Backdoor Mitigation by Unlearning Shared Adversarial Examples71
A novel upper bound for backdoor risk is derived, which mainly captures the risk on the shared adversarial examples (SAEs) between the backdoored model and the purified model, and a novel bi-level optimization problem for mitigating backdoor using adversarial training techniques is proposed.
- Robust Physical-World Attacks on Face Recognition71
A novel robust physical attack framework is proposed, dubbed PadvFace, to model these challenging variations of stickers, faces, and environmental conditions and an efficient Curriculum Adversarial Attack algorithm is proposed that gradually adapts adversarial stickers to environmental variations from easy to complex.
- Neural Polarizer: A Lightweight and Effective Backdoor Defense via Purifying Poisoned Features70
This work proposes a novel backdoor defense method by inserting a learnable neural polarizer into the backdoored model as an intermediate layer, in order to purify the poisoned sample via filtering trigger information while maintaining benign information.
- Improving Fast Adversarial Training With Prior-Guided Knowledge69
A positive prior-guided adversarial initialization to prevent overfitting by improving adversarial example quality without extra training time is proposed and can effectively improve the model's adversarial robustness in white-box attack scenarios.
- NOFA: NeRF-based One-shot Facial Avatar Reconstruction67
This work proposes a one-shot 3D facial avatar reconstruction framework that only requires a single source image to reconstruct a high-fidelity 3d facial avatar, and develops an efficient encoder-decoder network to reconstruct the canonical neural volume of the source image, and proposes a compensation network to complement facial details.
- Constrained Submodular Minimization for Missing Labels and Class Imbalance in Multi-label Learning62
This work forms the problem as a constrained submodular minimization that is composed of a sub modular objective function that encourages label consistency and smoothness, as well as, class cardinality bound constraints to handle class imbalance.
- Exact Adversarial Attack to Image Captioning via Structured Output Learning With Latent Variables61
This work proposes to fool an image captioning system to generate some targeted partial captions for an image polluted by adversarial noises, even the targeted captions are totally irrelevant to the image content.
- Prototype-supervised Adversarial Network for Targeted Attack of Deep Hashing57
A novel prototype-supervised adversarial network (ProS-GAN), which formulates a flexible generative architecture for efficient and effective targeted hashing attack and is the first generation-based method to attack deep hashing networks.
- Imperceptible and Robust Backdoor Attack in 3D Point Cloud55
This paper utilizes a nonlinear and local transformation, called weighted local transformation (WLT), to construct poisoned samples with unique transformations, which are imperceptible to human inspection and significantly higher than previous state-of-the-art attacks.
- Tagging Like Humans: Diverse and Distinct Image Annotation55
This work generates a relevant and distinct tag subset, in which the tags are relevant to the image contents and semantically distinct to each other, using sequential sampling from a determinantal point process (DPP) model.
- 53
- Orthogonal Subspace Decomposition for Generalizable AI-Generated Image Detection51
A novel approach to incorporate the rich semantic knowledge embedded within large-scale vision foundation models to expand the previous discriminative space (based on forgery patterns only), such that the discrimination is decided by both forgery and semantic cues, thereby reducing the overfitting to specific forgery patterns.
- Efficient Joint Gradient Based Attack Against SOR Defense for 3D Point Cloud Classification49
A novel white-box attack method, Joint Gradient Based Attack (JGBA), aiming to break the SOR defense, which generates adversarial examples by optimizing an objective function containing both the original point cloud and its SOR-processed version, for the purpose of pushing both of them towards the decision boundary of classifier at the same time.
- 48
- 46
- 42
- 42
- BlackboxBench: A Comprehensive Benchmark of Black-Box Adversarial Attacks41
This work aims at building a comprehensive benchmark of black-box adversarial attacks, implementing 29 query-based attack algorithms and 30 transfer-based attack algorithms and evaluating the implemented algorithms against several mainstreaming model architectures on 2 widely used datasets.
- 41
- Facial action unit recognition under incomplete data based on multi-label learning with missing labels40
Facial action unit recognition under incomplete data as a multi-label learning with missing labels (MLML) problem is formulated and the superiority of the proposed method in comparison with several state-of-the-art works is demonstrated.
- A Coupled Hidden Markov Random Field model for simultaneous face clustering and tracking in videos39
This paper introduces a Coupled Hidden Markov Random Field (CHMRF) to simultaneously model face clustering, face tracking, and their interactions, and provides an effective algorithm based on constrained clustering and optimal tracking for the joint optimization of cluster labels and face tracking.
- Breaking the False Sense of Security in Backdoor Defense through Re-Activation Attack38
This work re-investigate the characteristics of the backdoored models after defense and finds that the original backdoors still exist in defense models derived from existing post-training defense strategies, and the backdoor existence is measured by a novel metric called backdoor existence coefficient.
- 37
- Decentralized Directed Collaboration for Personalized Federated Learning36
A directed collaboration DPFL framework by incorporating stochastic gradient push and partial model personalized, called Decentralized Federated Partial Gradient Push (DFedPGP) is proposed, which achieves state-of-the-art (SOTA) accuracy in both data and computation heterogeneity scenarios.
- 36
- BackdoorBench: A Comprehensive Benchmark and Analysis of Backdoor Learning35
A comprehensive benchmark of backdoor learning is constructed, dubbed BackdoorBench, which provides an integrated implementation of representative backdoor learning algorithms, based on an extensible modular-based codebase, and reveals several inspiring insights about backdoor learning.
- 35
- Defenses in Adversarial Machine Learning: A Survey35
This survey aims to build a systematic review of all existing defense paradigms from a unified perspective, and factorize a complete machine learning system into five stages, including pre-training, training, post-training, deployment, and inference stages, respectively.
- 35
- 35
- A Survey of Trustworthy Graph Learning: Reliability, Explainability, and Privacy Protection33
This survey provides a comprehensive review of recent leading approaches in the TwGL field from three dimensions, namely, reliability, explainability, and privacy protection, and gives a general categorization for existing work and review typical work for each category.
- BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation32
This work introduces a comprehensive dataset, benchmark, and baseline model for video forgery detection, and develops BusterX, an MLLM baseline with RL training that outperforms several leading MLLMs in both detection accuracy and rationale quality.
- Visual Adversarial Attacks and Defenses in the Physical World: A Survey32
To establish a taxonomy, the current physical attacks from attack tasks, attack forms, and attack methods are organized so that readers can have a systematic knowledge about this topic from different aspects.
- 31
- Versatile Weight Attack via Flipping Limited Bits30
A novel attack paradigm, the bit-flip based weight attack, which directly modifies weight bits of the attacked model in the deployment stage and can be effectively and efficiently obtained by the alternating direction method of multipliers method.
- 29
- 27
- Visual Prompt Based Personalized Federated Learning23
A novel PFL framework for image classification tasks, dubbed pFedPT, is proposed that leverages personalized visual prompts to implicitly represent local data distribution information of clients and provides that information to the aggregation model to help with classification tasks.
- 22
- Backdoor Attack with Sample-Specific Triggers.22
Inspired by the recent advance in DNN-based image steganography, sample-specific invisible additive noises as backdoor triggers are generated by encoding an attacker-specified string into benign images through an encoder-decoder network.
- 22
- Data-Independent Operator: A Training-Free Artifact Representation Extractor for Generalizable Deepfake Detection21
This work shows that the small and training-free filter is sufficient to capture more general artifact representations and defines the DIO as Data-Independent Operator (DIO) to achieve appealing improvements on unseen sources.
- Activation Gradient based Poisoned Sample Detection Against Backdoor Attacks21
An innovative three-stage poisoned sample detection approach, called Activation Gradient based Poisoned sample Detection (AGPD), which filters out poisoned samples within the identified target class(es) based on the clear separation between poisoned and clean samples.
- Pixel‐wise Dense Detector for Image Inpainting21
A novel detection‐based generative framework for image inpainting, which adopts the min‐max strategy in an adversarial process and explicitly inserts the output of the detector into the reconstruction loss with a weighting criterion, which balances the weight of the adversarial loss and reconstruction loss automatically rather than manual operation.
- Learning to Learn from APIs: Black-Box Data-Free Meta-Learning20
A Bi-level Data-free Meta Knowledge Distillation (BiDf-MKD) framework to transfer more general meta knowledge from a collection of black-box APIs to one single meta model by just querying APIs is proposed.
- 20
- ToonTalker: Cross-Domain Face Reenactment18
This paper proposes a transformer-based framework to align the motions from different domains into a common latent space where motion transfer is conducted via latent code addition and proposes a novel cross-domain reenactment without paired data.
- RAIDX: A Retrieval-Augmented Generation and GRPO Reinforcement Learning Framework for Explainable Deepfake Detection17
RAIDX (Retrieval-Augmented Image Deepfake Detection and Explainability) is introduced, a novel deepfake detection framework integrating Retrieval-Augmented Generation and Group Relative Policy Optimization to enhance detection accuracy and decision explainability.
- 17
- 15
- Regional Adversarial Training for Better Robust Generalization14
A new adversarial training framework that considers the diversity as well as characteristics of the perturbed points in the vicinity of benign samples, and proposes a Regional Adversarial Training (RAT) defense method that first utilizes the attack path generated by the typical iterative attack method of projected gradient descent (PGD).
- 14
- To Think or Not to Think: Exploring the Unthinking Vulnerability in Large Reasoning Models13
A novel attack scenario targeting the long thought processes of o1-like models is introduced and BoT (Break CoT), which can selectively break intrinsic reasoning mechanisms through backdoor attacks is proposed, highlighting the critical safety risk in current models.
- Versatile Backdoor Attack With Visible, Semantic, Sample-Specific and Compatible Triggers13
By providing the first automated pipeline, VSSC transforms physical backdoor attacks from a labor-intensive craft into a systematic and realistic threat to real-world AI systems and could inspire future studies on designing more practical triggers in backdoor attacks.
- 13
- 13
- 13
- Attacks in Adversarial Machine Learning: A Systematic Survey from the Lifecycle Perspective12
It is demonstrated that isolated defenses are insufficient against coordinated multi-stage attacks, and the research community must prioritize holistic defense strategies incorporating lifecycle-aware monitoring, adaptive hardening techniques, and unified threat models.
- BadVideo: Stealthy Backdoor Attack Against Text-to-Video Generation12
This work reveals the adversarial vulnerability of T2V generative models, calling attention to potential risks and misuse, and introduces BadVideo, the first backdoor attack framework tailored for T2V generation.
- 12
- Self-Supervised AI-Generated Image Detection: A Camera Metadata Perspective11
Extensive experiments across various generative models demonstrate that the EXIF-induced detectors substantially advance the state of the art, delivering strong generalization to in-the-wild samples and robustness to common benign image perturbations.
- Towards Corruption-Agnostic Robust Domain Adaptation11
This work investigates a new task, corruption-agnostic robust domain adaptation (CRDA), to be accurate on original data and robust against unavailable-for-training corruptions on target domains and proposes an easy-to-plug module called domain discrepancy generator (DDG) that generates samples that enlarge domain discrepancy to mimic unpredictable corruptions.
- Handling missing labels and class imbalance challenges simultaneously for facial action unit recognition11
A convex approximation based on the Lovasz extension is presented, which leads to a linear program that can be efficiently solved by the alternative direction method of multipliers (ADMM) and demonstrates the superiority of the proposed method compared to state-of-the-art.
- A Coupled Hidden Conditional Random Field Model for Simultaneous Face Clustering and Naming in Videos11
This work divides the problem of face naming in videos when subtitles are not available into two tasks: face clustering which groups the faces depicting a certain person into a cluster, and name assignment which associates a name to each face.
- Hiding Faces in Plain Sight: Defending DeepFakes by Disrupting Face Detection10
A new proactive defense framework to prevent individuals from being victimized in DeepFake videos is investigated, which fools face detectors by adding dedicated adversarial perturbations to video frames and introduces VideoFacePoison, an extended strategy that can efficiently propagate FacePoison across video frames instead of applying it individually to each frame, thus significantly reducing the computational overhead while retaining favorable attack performance.
- Effective and Robust Detection of Adversarial Examples via Benford-Fourier Coefficients10
This work studies the detection of adversarial examples based on the assumption that the output and internal responses of one DNN model for both adversarial and benign examples follow the generalized Gaussian distribution but with different parameters (i.e., shape factor, mean, and variance).
- SPL-MLL: Selecting Predictable Landmarks for Multi-label Learning10
This work proposes to select a small subset of labels as landmarks which are easy to predict according to input (predictable) and can well recover the other possible labels (representative) to ensure both the representativeness and predictableness for selected landmarks.
- Enhanced Few-Shot Class-Incremental Learning via Ensemble Models9
A new ensemble model framework cooperated with data augmentation to boost generalization is designed, which works as a library storing abundant features to guarantee fast adaptation to downstream tasks and mitigate the overfitting problem in FSCIL.
- Adaptive Smoothness-weighted Adversarial Training for Multiple Perturbations with Its Stability Analysis8
The smoothness analysis of ATMP is provided and it is shown that adversaries give different contributions to the smoothness of the loss function of AT MP, which develops the stability-based excess risk bounds and proposes adaptive smoothness-weighted adversarial training for multiple perturbations.
- 8
- 7
- AdvChain: Adversarial Chain-of-Thought Tuning for Robust Safety Alignment of Large Reasoning Models6
AdvChain is proposed, an alignment paradigm that teaches models dynamic self-correction through adversarial CoT tuning that significantly enhances robustness against jailbreak attacks and CoT hijacking while substantially reducing over-refusal on benign prompts, achieving a superior safety-utility balance without compromising reasoning capabilities.
- Spurious Feature Eraser: Stabilizing Test-Time Adaptation for Vision-Language Foundation Model6
This work introduces a test-time prompt tuning paradigm that optimizes a learnable prompt, thereby compelling the model to exploit invariant features while disregarding decision shortcuts during the inference phase, and conducts comparative analysis of the proposed method against various approaches which validates the significant superiority.
- Robust and Minimally Invasive Watermarking for EaaS6
A novel embedding-specific watermarking (ESpeW) mechanism to offer robust copyright protection for EaaS and is minimally invasive, as it reduces the impact on embeddings to less than 1\%, setting a new milestone in watermarking for EaaS.
- Task-Distributionally Robust Data-Free Meta-Learning6
This paper proposes a trustworthy DFML framework comprising three components: synthetic task reconstruction, meta-learning with task memory interpolation, and automatic model selection, which seamlessly incorporates an automatic model selection mechanism to automatically filter out untrustworthy models during the meta-learning process.
- 6
- 6
- 6
- Efficient Black-Box Adversarial Attack Guided by the Distribution of Adversarial Perturbations.6
This work proposed to transform the Gaussian-distributed variable to another space through a conditional flow-based model, to enhance the capability and flexibility of capturing the intrinsic distribution of adversarial perturbations conditioned on the benign example, and pre-train the conditional flow model based on some white-box surrogate models.
- Groupwise Ranking Loss for Multi-Label Learning6
This work proposes a new loss, named Groupwise Ranking LosS (GRLS), and suggests that multi-label predictor is label-wise effective if and only if GRLS is optimal.
- 6
- The Authorization-Execution Gap Is a Major Safety and Security Problem in Open-World Agents5
It is argued that many observed agent failures can be traced to three structural sources of AEG: delegation-level incompleteness, channel-level corruption, and composition-level fragmentation, and that papers on open-world agents should report not only outcome-level metrics such as task success or attack resistance, but also process-level evidence showing where AEG was detected, constrained, and attributed to a structural source during execution.
- 5
- 5
- 5
- MAP Inference Via $$\ell _2$$-Sphere Linear Program Reformulation5
A perturbed alternating direction method of multipliers (ADMM) algorithm is proposed to optimize the LS–LP problem, by adding a sufficiently small perturbation of the original MAP inference problem, and it is proved that the perturbed ADMM algorithm globally converges to the ϵ\documentclass[12pt]{minimal} \usepackage{amsmath} \use package{wasysym} ℓ2.
- Constrained K-means with General Pairwise and Cardinality Constraints5
This paper enforces pairwise and cardinality constraints into a unified clustering model starting with the integer program formulation of the standard K-means and derives an equivalent continuous reformulation with simple constraints, which can be efficiently solved by Alternating Direction Method of Multipliers (ADMM).
- 4
- 4
- Reliable Poisoned Sample Detection against Backdoor Attacks Enhanced by Sharpness Aware Minimization4
This work proposes a novel approach that can complement existing detection methods and provide new insights about PSD and proposes a novel approach that can complement existing detection methods, which may inspire more in-depth explorations in this field.
- 4
- Learning to Accelerate Approximate Methods for Solving Integer Programming via Early Fixing4
The results demonstrate that the proposed innovative framework of learning to early fix variables along with the approximate method could not only significantly accelerate the previous approximate method up to over 10 times in most cases, but also produce similar or even better solutions.
- 3
- Fragile Model Watermark for integrity protection: leveraging boundary volatility and sensitive sample-pairing3
This work employs a sample-pairing technique, placing the model boundaries between pairs of samples, while simultaneously maximizing logits, which ensures that the model’s decision results of sensitive samples change as much as possible and the Top-1 labels easily alter regardless of the direction it moves.
- Teacher-Student Competition for Unsupervised Domain Adaptation3
This paper proposes an unsupervised domain adaptation approach with Teacher-Student Competition (TSC), in particular, a student network is introduced to learn the target-specific feature space, and a novel competition mechanism to select more credible pseudo-labels for the training of student network.
- 3
- Smart touchless palm sensing via palm adjustment and dynamic registration2
A comprehensive sensing system that integrates dynamic registration with robust palm adjustment is introduced, and an edge-aware, rotation-invariant region of interest alignment method is proposed that ensures consistent palm region localization.
- 2
- Class-Conditional Neural Polarizer: A Lightweight and Effective Backdoor Defense by Purifying Poisoned Features2
A lightweight backdoor defense approach to class-conditional neural polarizer-based defense, which inherently mimics targeted adversarial defense mechanisms without requiring label estimation used in NPD.
- 2
- 1
- 1
- 1
- HMGIE: Hierarchical and Multi-Grained Inconsistency Evaluation for Vision-Language Data Cleansing1
An adaptive evaluation framework, called Hierarchical and Multi-Grained Inconsistency Evaluation (HMGIE), which can provide multi-grained evaluations covering both accuracy and completeness for various image-caption pairs is designed and implemented.
- Inter-frame Accelerate Attack against Video Interpolation Models1
A novel attack method named Inter-frame Accelerate Attack (IAA) is proposed that initializes the perturbation as the perturgation for the previous adjacent frame and reduces the number of attack iterations and can improve attack efficiency greatly.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Interactive multimodal disease risk evaluation using a self-service non-invasive device–
iMES achieves robust performance across nine diseases with mean accuracy > 92% in a real-world cohort, indicating feasibility for scalable prevention-oriented screening and establishing iMES as a deployable solution for decentralized disease prevention in community and low-resource settings.
- –
- –
- –
- –
- Towards Explainable Bilingual Multimodal Misinformation Detection and Localization–
BiMi is introduced, a bilingual multimodal framework that jointly performs region-level localization, cross-modal and cross-lingual consistency detection, and natural language explanation for misinformation analysis and applies Group Relative Policy Optimization to improve explanation quality.
- –
- Reliable Imputed-Sample Assisted Vertical Federated Learning–
A Reliable Imputed-Sample Assisted VFL framework is proposed to effectively exploit non-overlapping samples by selecting reliable imputed samples for training VFL models and introduces evidence theory to estimate the uncertainty of imputed samples.
- –
- Customized Summarizations of Visual Data Collections–
The results show that the problem formulation enables a wide variety of customized summarizations, and that the solver is both significantly faster than state‐of‐the‐art commercial integer programming solvers and produces better solutions than fast relaxation‐based solvers.
- –
- –
- The Design of a New Instrument for Pen-contact Force Information Acquisition during Handwriting–
This paper describes a design and implementation of a new force-based instrument for the measurement of pen-contact force information that has the ability to detect 3-D forces and 2-D torques of writing force.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.