Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks76 from public data
- Spatial-Phase Shallow Learning: Rethinking Face Forgery Detection in Frequency Domain629
A novel Spatial-Phase Shallow Learning (SPSL) method is presented, which combines spatial image and phase spectrum to capture the up-sampling artifacts of face forgery to improve the transferability and theoretically analyze the validity of utilizing the phase spectrum.
- Towards Robust Vision Transformer263
Robust Vision Transformer (RVT) is proposed, which is a new vision transformer and has superior performance with strong robustness and generalization ability compared with previous ViTs and state-of-the-art CNNs.
- One-dimensional Adapter to Rule Them All: Concepts, Diffusion Models and Erasing Applications202
The concept-SemiPermeable structure is injected as a Membrane (SPM) into any DM to learn targeted erasing, and mean-time the alteration and erosion phenomenon is effectively mitigated via a novel Latent Anchoring fine-tuning strategy.
- SpanMlt: A Span-based Multi-Task Learning Framework for Pair-wise Aspect and Opinion Terms Extraction185
This paper proposes an end-to-end method to solve the task of Pair-wise Aspect and Opinion Terms Extraction (PAOTE), and treats the problem from a perspective of joint term and relation extraction rather than under the sequence tagging formulation performed in most prior works.
- Self-Supervised Learning for Few-Shot Image Classification134
This paper proposed to train a more generalized embedding network with self-supervised learning (SSL) which can provide robust representation for downstream tasks by learning from the data itself.
- Towards Face Encryption by Generating Adversarial Identity Masks127
This work proposes a targeted identity-protection iterative method (TIP-IM) to generate adversarial identity masks which can be overlaid on facial images, such that the original identities can be concealed without sacrificing the visual quality.
- Large Language Models Can be Lazy Learners: Analyze Shortcuts in In-Context Learning108
It is revealed that large language models are "lazy learners" that tend to exploit shortcuts in prompts for downstream tasks and a surprising finding is uncovered that larger models are more likely to utilize shortcuts in Prompts during inference.
- Towards discovering and understanding task hijacking in android102
Design flaws of Android multitasking are found which make all recent versions of Android vulnerable to task hijacking attacks, and possible mitigation techniques are discussed.
- 95
- Synthesizing Coherent Story with Auto-Regressive Latent Diffusion Models91
This work proposes AR-LDM, a latent diffusion model auto-regressively conditioned on history captions and generated images that can generalize to new characters through adaptation and extends the text-conditioned method to multimodal conditioning.
- Composite Adversarial Attacks61
A new procedure called Composite Adversarial Attack (CAA) is proposed for automatically searching the best combination of attack algorithms and their hyper-parameters from a candidate pool of 32 base attackers, and multi-objective NSGA-II genetic algorithm is adopted for finding the strongest attack policy with minimum complexity.
- Fine-Grained Fashion Similarity Learning by Attribute-Specific Embedding Network61
An Attribute-Specific Embedding Network (ASEN) is proposed to jointly learn multiple attribute-specific embeddings in an end-to-end manner, thus measure the fine-grained similarity in the corresponding space.
- Verifying security invariants in ExpressOS61
The evaluation shows that the performance of ExpressOS is comparable to an Android-based system, and it is found that ExpressOS adds 16% overhead on average to the page load latency time for nine popular web sites.
- QAIR: Practical Query-efficient Black-Box Attacks for Image Retrieval54
This paper makes the first attempt in Query-based Attack against Image Retrieval (QAIR), to completely subvert the top-k retrieval results by measuring the set similarity on theTop-K retrieval results before and after attacks and guide the gradient optimization.
- 45
- Context-Aware Robust Fine-Tuning44
Context-Aware Robust Fine-tuning (CAR-FT) makes the context-aware ability of CLIP inherited into downstream tasks, and achieves both higher in-distribution (ID) and out-of-distribution (OOD) accuracy.
- ImageNet-E: Benchmarking Neural Network Robustness via Attribute Editing43
This paper conducts model debugging in in-distribution data to explore which object attributes a model may be sensitive to, and develops a toolkit for object editing with controls of backgrounds, sizes, positions, and directions, and creates a rigorous benchmark named ImageNet-E(diting) for evaluating the image classifier robustness in terms of object attributes.
- Fighting Against Deepfake: Patch&Pair Convolutional Neural Networks (PPCNN)43
Through the comprehensive evaluations on public datasets, it is demonstrated that the proposed PPCNN model performs better than existing detection methods and show better generalization.
- S-Eval: Towards Automated and Comprehensive Safety Evaluation for Large Language Models38
S-Eval is a novel LLM-based automated Safety Evaluation framework with a newly defined comprehensive risk taxonomy that can be flexibly configured and adapted to the rapid evolution of LLMs and accompanying new safety threats, test generation methods and safety critique methods thanks to the LLM-based architecture.
- Dynamic Mixture of Curriculum LoRA Experts for Continual Multimodal Instruction Tuning33
A novel Dynamic Mixture of Curriculum LoRA Experts (D-MoLE) method, which automatically evolves MLLM's architecture with controlled parameter budgets to continually adapt to new tasks while retaining previously learned knowledge, is proposed, the first study of continual learning for MLLMs from an architectural perspective.
- 31
- 26
- 26
- Self-Supervised Adversarial Training26
Experimental results show that the self- supervised representation outperforms its supervised version in respect of robustness and self-supervised adversarial training can further improve the defense ability efficiently.
- Bilinear Representation for Language-based Image Editing Using Conditional Generative Adversarial Networks25
This work proposes an improved conditional layer named Bilinear Residual Layer (BRL) to learning more powerful representations for LBIE task and demonstrates that this method can generate images with higher quality when compared to previous LBIe techniques.
- Revisiting and Exploring Efficient Fast Adversarial Training via LAW: Lipschitz Regularization and Auto Weight Averaging21
An effective FGSM-based fast adversarial training method equipped with Lipschitz regularization and Auto Weight averaging, abbreviated as FGSM-LAW is proposed, demonstrating the superiority of this method over state-of-the-art fast adversarial training methods and the advanced standard adversarial training methods.
- 21
- The Open Brands Dataset: Unified Brand Detection and Recognition at Scale21
A novel brands benchmark called "Open Brands" is established, a network called "Brand Net" is designed to handle brand recognition and Brand Net gets state-of-art mAP on Open Brand compared with existing detection methods.
- Differential-Perceptive and Retrieval-Augmented MLLM for Change Captioning20
This work aims to empower MLLMs with the capability to perceive subtle differences between paired images and enhance their performance in generating change captions by presenting a diFferental-perceptive, medieval-augmented MLLM (FINER-MLLM) tailored for this task.
- Smart Fire Alarm System with Person Detection and Thermal Camera20
This paper introduces a smart fire alarm system used in kitchen as a representative scenario, which captures both thermal and optical videos for temperature monitoring and person detection and is further used to predict potential fire accident and avoid false alarm.
- Using replicated execution for a more secure and reliable web browser.18
It is demonstrated that Cocktail can withstand real-world browser exploits and reliability issues, such as browser crashes, while adding only 31.5% overhead to page load latency times on average, and remaining compatible with popular web sites.
- BiFPro: A Bidirectional Facial-data Protection Framework against DeepFake17
Experimental results demonstrate that the proposed Bidirectional Facial-data Protection Framework could generate the watermarking which is fragile to FOD and at the same time robust to FIF with an average watermark extraction success rate reaching more than 95% when defending against the four advanced DeepFake techniques.
- Spatio-Temporal Catcher: A Self-Supervised Transformer for Deepfake Video Detection17
It is argued that data quality has become the main bottleneck of current research and a novel spatio-temporal generator that can synthesize various high-quality "fake" videos in large quantities at a low cost is developed, which enables the model to learn more general spatio/temporal representations in a self-supervised manner.
- Model Inversion Attack via Dynamic Memory Learning16
This work proposes a novel Dynamic Memory Model Inversion Attack (DMMIA) to leverage historically learned knowledge, which interacts with samples (during the training) to induce diverse generations and performs better than state-of-the-art MI attack methods.
- On the Role of Long-tail Knowledge in Retrieval Augmented Large Language Models15
It is suggested that long-tail knowledge is crucial for RAG as LLMs have already remembered common world knowledge during large-scale pre-training and can be used for patching knowledge loopholes only when the input query relates to long-tail knowledge.
- Incremental Graph Classification by Class Prototype Construction and Augmentation15
This work proposes a novel method for class-incremental graph learning (CGL) by class prototype construction and augmentation, which can effectively overcome catastrophic forgetting and requires no storage of exemplars (i.e., data-free).
- Continual Few-shot Learning with Transformer Adaptation and Knowledge Regularization14
A Continual few-shot learning algorithm with Semantic knowledge Regularization (CoSR) for adapting to the distribution changes of visual prototypes through a Transformer-based prototype adaptation mechanism, which improves the model’s ability to distinguish between new and old categories, thus significantly mitigating the catastrophic forgetting problem in continual few- shot learning.
- Learning to Characterize Adversarial Subspaces13
This work proposes a novel adversarial detection method which identifies adversaries by adaptively learning reasonable metrics to characterize adversarial subspaces and proposes an innovative model called Neighbor Context Encoder (NCE) to learn from k neighbors context and infer if the detected sample is normal or adversarial.
- Large Language Model with Curriculum Reasoning for Visual Concept Recognition12
A novel rule enhancement module with a tool library, which fully leverage the reasoning capability of large language models and can generate human-understandable rules without any annotation, and a curriculum data resampling methodology to help the large language model accurately extract from easy to complex rules at different reasoning stages.
- Enhancing Model Robustness by Incorporating Adversarial Knowledge into Semantic Representation12
AdvGraph is proposed, a novel defense which enhances the robustness of Chinese-based NLP models by incorporating adversarial knowledge into the semantic representation of the input and is a light-weight defense with sub-linear computational complexity.
- A Visual Inductive Priors Framework for Data-Efficient Image Classification12
A novel neural network architecture: DSK-net is proposed, which is very effiective in training from small data sets and with more discriminative feature extracted from DSK-net, overfitting of network is alleviated.
- Accurate and efficient exploit capture and classification12
SeismoMeter is presented, which recognizes both control-flowhijacking, and data-only attacks by combining approximate control-flow integrity, fast dynamic taint analysis and API sandboxing schemes and is demonstrated to be a practical system that successfully detects and correctly classifies all these exploit attacks.
- 10
- Rethinking Out-of-Distribution Detection From a Human-Centric Perspective10
A human-centric evaluation of OOD detection is proposed and it is found that the simple baseline OOD detection method can achieve comparable and even better performance than the recently proposed methods, which means that the development in OOD detection in the past years may be overestimated.
- 10
- 9
- Improving Model Generalization by On-manifold Adversarial Augmentation in the Frequency Domain6
Extensive experiments on different models and different datasets, including ImageNet and its distorted versions, demonstrate that the proposed novel method of Augmenting data with Adversarial examples via a Wavelet module (AdvWavAug) can improve model generalization, especially on OOD data.
- 6
- 6
- 5
- 5
- Generative Video Matting4
This work introduces a novel video matting approach that can effectively leverage the rich priors from pre-trained video diffusion models, and offers two key advantages, unlike most existing methods that process video matting frame-by-frame and use an independent decoder to aggregate temporal information.
- Towards the Resistance of Neural Network Watermarking to Fine-tuning4
It is proved that when the input feature of a convolutional layer only contains low-frequency components, specific frequency components of the convolutional filter will not be changed by gradient descent during the fine-tuning process, where a revised Fourier transform is proposed to extract frequency components from the convolutional filter.
- Score-based Generative Modeling for Conditional Independence Testing4
This work proposes a novel CI testing method via score-based generative modeling, which achieves precise Type I error control and strong testing power, and incorporates a goodness-of-fit stage into the method to verify generated samples and enhance interpretability in practice.
- 4
- 4
- 4
- 4
- Learning from Graph: Mitigating Label Noise on Graph through Topological Feature Reconstruction3
This work theoretically and empirically demonstrates that the graph itself contains reliable information for graph learning under label noise, and proposes the Topological Feature Reconstruction (TFR) method, a simple and theoretically guaranteed model for robust graph learning under label noise.
- 2
- Diverse Instance Discovery: Vision-Transformer for Instance-Aware Multi-Label Image Recognition2
This paper makes full use of the advantages of Transformer with long-range dependency modeling to circumvent the disadvantages of CNNs limited to local receptive field and proposes a weakly supervised object localization-based approach to extract multi-scale local features, to form a multi-view pipeline.
- 2
- 2
- Learning to Detect Unseen Jailbreak Attacks in Large Vision-Language Models1
This work proposes Learning to Detect (LoD), a general framework that accurately detects unknown jailbreak attacks by shifting the focus from attack-specific learning to task-specific learning and includes a Multi-modal Safety Concept Activation Vector module for safety-oriented representation learning.
- 1
- 1
- 1
- 1
- –
- A Comprehensive Toolkit for Generalized Robust Vision–
A comprehensive, user-friendly toolkit for training, evaluating, and analyzing robust vision models that targets two key dimensions of robustness: Adversarial robustness-defending against malicious worst-case perturbations (adversarial examples) and Natural robustness-maintaining performance under real-world corruptions and distribution shifts.
- –
- –
- –
- PIAT: Parameter Interpolation based Adversarial Training for Image Classification–
A novel framework, termed Parameter Interpolation based Adversarial Training (PIAT), that makes full use of the historical information during training and tunes the model parameters as the interpolation of the parameters of the previous and current epochs.
- –
- DRDF–
This work proposes a high-performance and highly general Dual-Router Dynamic Framework (DRDF), consisting of dual router, MWF-Layer, experts and expert fusion unit, and test 12 backbones such as Visual BERT and their corresponding DRDF instances on the multimodal dataset Hateful memes, and unimodal datasets CifAR10, CIFAR100, and TinyImagenet.
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 2026-10-14. 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.