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
- James BaileySuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works171 from public data
- Symmetric Cross Entropy for Robust Learning With Noisy Labels1,212
The proposed Symmetric cross entropy Learning (SL) approach simultaneously addresses both the under learning and overfitting problem of CE in the presence of noisy labels, and empirically shows that SL outperforms state-of-the-art methods.
- Improving Adversarial Robustness Requires Revisiting Misclassified Examples876
This paper proposes a new defense algorithm called MART, which explicitly differentiates the misclassified and correctly classified examples during the training, and shows that MART and its variant could significantly improve the state-of-the-art adversarial robustness.
- Reflection Backdoor: A Natural Backdoor Attack on Deep Neural Networks640
Refool is proposed, a new type of backdoor attack inspired by an important natural phenomenon: reflection to plant reflections as backdoor into a victim model and can attack state-of-the-art DNNs with high success rate, and is resistant to state of theart backdoor defenses.
- WildDeepfake609
WildDeepfake is a small dataset that can be used, in addition to existing datasets, to develop and test the effectiveness of deepfake detectors against real-world deepfakes, and proposes two (eg. 2D and 3D) Attention-based Deepfake Detection Networks (ADDNets) to leverage the attention masks on real/fake faces for improved detection.
- Privacy and Robustness in Federated Learning: Attacks and Defenses582
A comprehensive survey on privacy and robustness in FL over the past five years is conducted and a concise introduction to the concept of FL is provided and a unique taxonomy covering privacy attacks and defenses is provided.
- Understanding adversarial attacks on deep learning based medical image analysis systems567
It is found that medical DNN models can be more vulnerable to adversarial attacks compared to models for natural images, according to two different viewpoints, and surprisingly, medical adversarialattacks can be easily detected.
- 483
- Iterative Learning with Open-set Noisy Labels361
A novel iterative learning framework for training CNNs on datasets with open-set noisy labels that detects noisy labels and learns deep discriminative features in an iterative fashion and designs a Siamese network to encourage clean labels and noisy labels to be dissimilar.
- Clean-Label Backdoor Attacks on Video Recognition Models335
This paper proposes the use of a universal adversarial trigger as the backdoor trigger to attack video recognition models, a situation where backdoor attacks are likely to be challenged by the above 4 strict conditions.
- Adversarial Camouflage: Hiding Physical-World Attacks With Natural Styles285
Experimental evaluation shows that, in both digital and physical-world scenarios, adversarial examples crafted by Adversarial Camouflage are well camouflaged and highly stealthy, while remaining effective in fooling state-of-the-art DNN image classifiers.
- 255
- 238
- 182
- Black-box Adversarial Attacks on Video Recognition Models175
V-BAD is a promising new tool to evaluate and improve the robustness of video recognition models to black-box adversarial attacks, and can craft both untargeted and targeted attacks to fool two state-of-the-art deepVideo recognition models.
- 173
- Revisiting Adversarial Robustness Distillation: Robust Soft Labels Make Student Better150
The effectiveness of the RSLAD approach over existing adversarial training and distillation methods in improving the robustness of small models against state-of-the-art attacks including the AutoAttack is empirically demonstrated.
- A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deployment139
This paper introduces the concept of "full-stack" safety to systematically consider safety issues throughout the entire process of LLM training, deployment, and eventual commercialization, and represents the first safety survey to encompass the entire lifecycle of LLMs.
- 139
- 121
- Reconstructive Neuron Pruning for Backdoor Defense107
This paper proposes a novel defense called Reconstructive Neuron Pruning (RNP) to expose and prune backdoor neurons via an unlearning and then recovering process and shows that such an asymmetric process on only a few clean samples can effectively expose andPrune the backdoor neurons implanted by a wide range of attacks.
- Copy, Right? A Testing Framework for Copyright Protection of Deep Learning Models107
A novel testing framework for deep learning copyright protection: DEEPJUDGE quantitatively tests the similarities between two deep learning models: a victim model and a suspect model, which leverages a diverse set of testing metrics and efficient test case generation algorithms to produce a chain of supporting evidence to help determine whether a suspects model is a copy of the victim model.
- VeriFi: Towards Verifiable Federated Unlearning92
This work prompts the concept of verifiable federated unlearning and proposes VeriFi, a unified framework that allows systematic analysis of federated unlearning and quantification of its effect, with different combinations of various unlearning and verification methods.
- RobOT: Robustness-Oriented Testing for Deep Learning Systems87
Experiments on multiple benchmark datasets confirm the effectiveness and efficiency of RobOT in improving DL model robustness, with 67.02% increase on the adversarial robustness that is 50.65% higher than the state-of-the-art work DeepGini.
- White-box Multimodal Jailbreaks Against Large Vision-Language Models82
A universal attack strategy that jointly attacks both text and image modalities to exploit a broader spectrum of vulnerability within VLMs and can effectively jailbreak MiniGPT-4 with a 96% success rate, highlighting the fragility of VLMs and the exigency for new alignment strategies.
- 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.
- A Survey of Multimodal-Guided Image Editing with Text-to-Image Diffusion Models73
A comprehensive review of multimodal-guided image editing techniques that leverage T2I diffusion models, and proposes a unified framework to formalize the editing process, categorizing it into two primary algorithm families.
- Transferable Unlearnable Examples64
A novel unlearnable strategy based on Classwise Separability Discriminant (CSD) is proposed, which aims to better transfer the Unlearnable effects to other training settings and datasets by enhancing the linear separability.
- Short-Term and Long-Term Context Aggregation Network for Video Inpainting58
This work presents a novel context aggregation network to effectively exploit both short-term and long-term frame information for video inpainting, and outperforms state-of-the-art methods with better in painting results and fast inPainting speed.
- 57
- How to Democratise and Protect AI: Fair and Differentially Private Decentralised Deep Learning54
A fair and differentially private decentralised deep learning framework called FDPDDL is built, which enables parties to derive more accurate local models in a fair and private manner by using a developed two-stage scheme.
- CalFAT: Calibrated Federated Adversarial Training with Label Skewness54
This paper studies the problem of FAT under label skewness, and proposes a Calibrated FAT (CalFAT) approach to tackle the instability issue by calibrating the logits adaptively to balance the classes and shows that the optimization of CalFAT leads to homogeneous local models across the clients and better convergence points.
- Anyattack: Towards Large-scale Self-supervised Adversarial Attacks on Vision-language Models52
AnyAttack, a self-supervised framework that transcends the limitations of conventional attacks through a novel foundation model approach, is presented, enabling any image to be transformed into an attack vector targeting any desired output across different VLMs.
- BackdoorLLM: A Comprehensive Benchmark for Backdoor Attacks and Defenses on Large Language Models49
BackdoorLLM is the first comprehensive benchmark for systematically evaluating backdoor threats in text-generation LLMs and provides a diverse suite of attack modalities, including data poisoning, weight poisoning, hidden-state manipulation, and chain-of-thought hijacking.
- Unlearnable Clusters: Towards Label-Agnostic Unlearnable Examples49
This work presents a novel technique called Unlearnable Clusters (UCs) to generate label-agnostic unlearnable examples with cluster-wise perturbations and proposes to leverage Vision-and-Language Pre-trained Models (VLPMs) like CLIP as the surrogate model to improve the transferability of the crafted UCs to diverse domains.
- Machine learning guided alloy design of high-temperature NiTiHf shape memory alloys47
Of the studied models, the K-nearest neighbouring ML model offers more reliable and accurate prediction in developing NiTiHf alloys with balanced functional properties and aids the existing understanding on compositional dependence of transformation temperature, thermal hysteresis and work output.
- 44
- Distilling Cognitive Backdoor Patterns within an Image43
Despite the various forms and sizes of trigger patterns used by different attacks, the CPs of backdoor samples are all surprisingly and suspiciously small, so one can leverage the learned mask to detect and remove backdoor examples from poisoned training datasets.
- Toward Generalizable Evaluation in the LLM Era: A Survey Beyond Benchmarks39
This survey probes the core challenges that the rise of LLMs poses for evaluation, and identifies and analyzes two pivotal transitions: from task-specific to capability-based evaluation and from manual to automated evaluation.
- TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models38
TAPT is a test-time defense method that learns defensive bimodal (textual and visual) prompts to robustify the inference process of CLIP, and outperforms existing adversarial prompt tuning methods across various backbones.
- 37
- BlueSuffix: Reinforced Blue Teaming for Vision-Language Models Against Jailbreak Attacks35
This paper proposes a novel blue-team method BlueSuffix that defends target VLMs against jailbreak attacks without compromising its performance under black-box setting and empirically shows that BlueSuffix outperforms the baseline defenses by a significant margin.
- 35
- X-Transfer Attacks: Towards Super Transferable Adversarial Attacks on CLIP32
X-Transfer is introduced, a novel attack method that exposes a universal adversarial vulnerability in CLIP and generates a Universal Adversarial Perturbation (UAP) capable of deceiving various CLIP encoders and downstream VLMs across different samples, tasks, and domains.
- 31
- 31
- Detecting Backdoor Samples in Contrastive Language Image Pretraining26
This work analyzes the representations of backdoor-poisoned samples learned by CLIP models and finds that they exhibit unique characteristics in their local subspace, i.e., their local neighborhoods are far more sparse than that of clean samples.
- Shortcuts Everywhere and Nowhere: Exploring Multi-Trigger Backdoor Attacks25
This study explores the concept of Multi-Trigger Backdoor Attacks (MTBAs), where multiple adversaries leverage different types of triggers to poison the same dataset, and demonstrates that multiple triggers can coexist, overwrite, or cross-activate one another.
- Towards Fair and Decentralized Privacy-Preserving Deep Learning with Blockchain25
The proposed DPPDL framework provides a viable solution to detect and isolate the cheating party in the system, and simultaneously provides fairness and privacy by proposing two novel algorithms: initial benchmarking and privacy-preserving collaborative deep learning.
- Unbiased Multivariate Correlation Analysis25
UMC is a cumulative entropy based non-parametric multivariate correlation measure, which can capture both linear and non-linear correlations for groups of three or more variables, and employs a correction for chance using a statistical model of independence to address the issue of bias.
- Design and Evaluation of a Virtual Reality Simulation Module for Training Advanced Temporal Bone Surgery25
This paper discusses the design and evaluation of such a training module, specifically aimed at training an advanced temporal bone procedure, namely cochlear implant surgery.
- Learning from Heterogeneity: A Dynamic Learning Framework for Hypergraphs23
A hypergraph learning framework named learning from heterogeneity (LFH) that is capable of dynamic hyperedge construction and attentive embedding update utilizing the heterogeneity attributes of the graph is proposed.
- Backdoor Attacks on Time Series: A Generative Approach23
A novel generative approach for time series backdoor attacks against deep learning based time series classifiers and a novel universal generator that can poison any type of time series with a single generator that allows universal attacks without the need to fine-tune the generative model for new time series datasets.
- Providing Automated Real-Time Technical Feedback for Virtual Reality Based Surgical Training: Is the Simpler the Better?22
It is shown that simpler rule-based methods can be equally or more effective in teaching surgical skills when compared to more complex methods of feedback generation.
- On the Importance of Spatial Relations for Few-shot Action Recognition21
It is found that the spatial misalignment between objects also occurs in videos, notably more common than the temporal inconsistency, and a novel Spatial Alignment Cross Transformer (SA-CT) which learns to re-adjust the spatial relations and incorporates the temporal information is contributed.
- QuoTe : Quality-oriented Testing for Deep Learning Systems21
Extensive experiments on both image and tabular datasets with a variety of model architectures confirm the effectiveness and efficiency of QuoTe in improving DL model quality—that is, robustness and fairness.
- 21
- Special Characters Attack: Toward Scalable Training Data Extraction From Large Language Models20
This work shows that certain special characters or their combinations with English letters are stronger memory triggers, leading to more severe data leakage in large language models, and proposes a simple but effective Special Characters Attack (SCA) to induce training data leakage.
- Noise Doesn't Lie: Towards Universal Detection of Deep Inpainting20
This paper proposes a novel data generation approach to generate a universal training dataset, which imitates the noise discrepancies exist in real versus inpainted image contents to train universal detectors, and designs a Noise-Image Cross-fusion Network (NIX-Net) to effectively exploit the discriminative information contained in both the images and their noise patterns.
- Imbalanced Gradients: A New Cause of Overestimated Adversarial Robustness20
12 state-of-the-art defense models are examined, and it is found that models exploiting label smoothing easily cause imbalanced gradients, and on which attacks can decrease their PGD robustness by over 23%.
- 19
- Anomaly Detection for Scenario-based Insider Activities using CGAN Augmented Data19
A Conditional Generative Adversarial Network (CGAN) is proposed to enrich under-represented minority class samples to provide meaningful and diverse data for anomaly detection from the original malicious scenarios.
- Exploiting patterns to explain individual predictions18
PALEX is compared to several state-of-the-art explanation methods over a range of benchmark datasets and finds that it can identify explanations with both high precision and high recall.
- Generative Image Inpainting with Submanifold Alignment18
This work exploits Local Intrinsic Dimensionality (LID) to measure the alignment between data submanifolds learned by a GAN model and those of the original data, from a perspective of both images and local patches of images to enforce the alignment (or closeness) around restored images and those around the original (uncorrupted) images during the learning process of GAN-based inpainting models.
- 17
- Backdoor Attacks on Crowd Counting17
This paper proposes two novel Density Manipulation Backdoor Attacks (DMBA- and DMBA+) to attack the model to produce arbitrarily large or small density estimations, and provides an in-depth analysis of the unique challenges of backdooring crowd counting models.
- Imbalanced gradients: a subtle cause of overestimated adversarial robustness16
A margin decomposition (MD) attack that decomposes a margin loss into individual terms and then explores the attackability of these terms separately via a two-stage process to exploit imbalanced gradients.
- Free-Form Motion Control: Controlling the 6D Poses of Camera and Objects in Video Generation15
A Synthetic Dataset for Free-Form Motion Control (SynFMC) is introduced, which can control the 6D poses of objects and camera independently or simultaneously, producing high-fidelity videos and outperforms previous methods across multiple scenarios.
- LDReg: Local Dimensionality Regularized Self-Supervised Learning15
By increasing the local intrinsic dimensionality, it is demonstrated through a range of experiments that LDReg improves the representation quality of SSL and the results show that LDReg can regularize dimensionality at both local and global levels.
- 15
- 14
- AdvQDet: Detecting Query-Based Adversarial Attacks with Adversarial Contrastive Prompt Tuning14
A novel Adversarial Contrastive Prompt Tuning (ACPT) method is proposed to robustly fine-tune the CLIP image encoder to extract similar embeddings for any two intermediate adversarial queries and a detection framework AdvQDet is introduced that can detect 7 state-of-the-art query-based attacks with >99% detection rate within 5 shots.
- 14
- AttackVLA: Benchmarking Adversarial and Backdoor Attacks on Vision-Language-Action Models13
BackdoorVLA is introduced, a targeted backdoor attack that compels a VLA to execute an attacker-specified long-horizon action sequence whenever a trigger is present, providing a standardized framework for evaluating VLA vulnerabilities and demonstrating the potential for precise adversarial manipulation.
- Adversarial Generation of Real-time Feedback with Neural Networks for Simulation-based Training13
A neural network based method to generate feedback using the adversarial technique that was observed to have high effectiveness and efficiency when compared to existing methods, thus making it a promising option for real-time feedback generation in SBT.
- Query-Efficient Black-Box Adversarial Attacks on Automatic Speech Recognition12
A new gradient estimation technique, Temporal Natural Evolution Strategies (T-NES), is introduced, to generate adversarial audio samples more efficiently than existing attacks, and could provide a robust baseline for evaluating the black-box adversarial vulnerability of ASR systems.
- Development and Validation of a Virtual Reality Tutor to Teach Clinically Oriented Surgical Anatomy of the Ear11
The design and development of a 3D interactive VR tutor with integrated haptic capability to teach clinically oriented surgical anatomy of the ear is discussed and its effectiveness is established through a user study.
- IDEATOR: Jailbreaking and Benchmarking Large Vision-Language Models Using Themselves10
A novel jailbreak method named IDEATOR, which autonomously generates malicious image-text pairs for black-box jailbreak attacks and uncovers specific vulnerabilities in VLMs under black-box conditions, under-scoring the need for improved safety mechanisms.
- Providing Effective Real-Time Feedback in Simulation-Based Surgical Training10
A random forest based method is proposed that finds a balance between effectiveness and efficiency and is able to extract highly effective feedback at a high level of efficiency in a temporal bone surgery simulation.
- CHASE: A Causal Hypergraph based Framework for Root Cause Analysis in Multimodal Microservice Systems9
A Causal Heterogeneous framework for root cause analysis, namely CHASE, for microservice systems with multimodal data, including traces, logs, and system monitoring metrics is proposed and evaluated on two public microservice datasets with distinct attributes.
- FedCAda: Adaptive Client-Side Optimization for Accelerated and Stable Federated Learning9
FedCAda is introduced, an adaptive algorithm that leverages an Adam-like approach to adjust first and second moment estimates on the client side while aggregating adaptive parameters on the server side, to accelerate convergence without compromising stability and performance.
- DAVID-XR1: Detecting AI-Generated Videos with Explainable Reasoning9
DAMID-X is introduced, the first dataset to pair AI-generated videos with detailed defect-level, temporal-spatial annotations and written rationales, and DAVID-XR1, a video-language model designed to deliver an interpretable chain of visual reasoning-including defect categorization, temporal-spatial localization, and natural language explanations.
- Fuse Your Latents: Video Editing with Multi-source Latent Diffusion Models9
This paper is the first to reveal that T2I and T2V LDMs can complement each other in terms of structure and temporal consistency, ultimately generating high-quality videos.
- 9
- Federated Learning with Extreme Label Skew: A Data Extension Approach9
This work shows how using some wild external data samples as placeholders for missing classes on client devices can alleviate the learning difficulty often posed by inbalance data distributions, and shows that this strategy enhances learning and can significantly boost test accuracy.
- Simulation for Training Cochlear Implant Electrode Insertion9
A prototype implementation for the visualisation and analysis of electrode trajectories in relation to anatomical variation is discussed, prior to its inclusion in a virtual reality training module for cochlear implant surgery.
- ModelLock: Locking Your Model With a Spell8
A diffusion-based framework dubbed ModelLock is proposed that explores text-guided image editing to transform the private finetuning data into unique styles or blend new objects into the background to open up a new direction for intellectual property protection of private models.
- ECG-ATK-GAN: Robustness Against Adversarial Attacks on ECGs Using Conditional Generative Adversarial Networks8
The first novel Conditional Generative Adversarial Network (GAN) is introduced, robust against adversarial attacked ECG signals and retaining high accuracy, and is benchmarked on six different white and black-box attacks and confirms that the model is more robust against such adversarial attacks for classifying arrhythmia with high accuracy.
- Dual Head Adversarial Training8
This paper proposes a novel technique Dual Head Adversarial Training (DH-AT), which modifies both the architecture of the network and the training strategy to seek more robustness, and empirically shows that it can bring notable robustness improvements to existing adversarial training methods.
- SpineOne: A One-Stage Detection Framework for Degenerative Discs and Vertebrae8
This work proposes a one-stage detection framework termed SpineOne to simultaneously localize and classify degenerative discs and vertebrae from magnetic resonance imaging (MRI) slices and demonstrates that this approach surpasses existing methods by a large margin.
- 8
- AutoBackdoor: Automating Backdoor Attacks via LLM Agents7
This work introduces AutoBackdoor, a general framework for automating backdoor injection, encompassing trigger generation, poisoned data construction, and model fine-tuning via an autonomous agent-driven pipeline, and demonstrates that the method achieves over 90% attack success with only a small number of poisoned samples.
- SIDE: Surrogate Conditional Data Extraction from Diffusion Models7
A theoretical understanding of memorization in both conditional and unconditional DPMs under the assumption of model convergence is provided and a novel data extraction method named Surrogate condItional Data Extraction (SIDE) is proposed that leverages a time-dependent classifier trained on the generated data as a surrogate condition to extract training data from unconditional DPMs.
- Adversarial Prompt Distillation for Vision-Language Models7
Adversarial Prompt Distillation (APD) is proposed, a bimodal knowledge distillation framework that enhances APT by integrating it with multi-modal knowledge transfer and validates the possibility of using a non-robust teacher to improve the generalization and robustness of fine-tuned VLMs.
- Neural Architecture Search via Combinatorial Multi-Armed Bandit7
A tree-based method called Nested Monte-Carlo Search is used to tackle the CMAB-NAS problem and discovers a cell structure that achieves a low error rate that is comparable to the state of the art, using only 0.58 GPU days, which is 20 times faster than current tree-search methods.
- 7
- Surgical approach to the facial recess influences the acceptable trajectory of cochlear implantation electrodes7
Findings highlight the differences in the acceptable straight-line trajectories for electrodes when implanted via the round window or cochleostomy, and the methodology used holds promise for translation to patient specific surgical planning.
- Evolve the Method, Not the Prompts: Evolutionary Synthesis of Jailbreak Attacks on LLMs6
EvoSynth is an autonomous multi-agent framework that shifts the optimization space from prompts to executable code, and features a code-level self-correction loop, allowing it to iteratively rewrite the code-based algorithm in response to target-model feedback and failed attempts.
- 6
- GenBreak: Red Teaming Text-to-Image Generators Using Large Language Models5
GenBreak is a framework that fine-tunes a red-team large language model (LLM) to systematically explore underlying vulnerabilities in T2I generators, and guides the LLM to craft adversarial prompts that enhance both evasion capability and image toxicity, while maintaining semantic coherence and diversity.
- Imperceptible Jailbreaking against Large Language Models5
This paper introduces imperceptible jailbreaks that exploit a class of Unicode characters called variation selectors that appear visually identical to original malicious questions on screen, while their tokenization is"secretly"altered.
- EnJa: Ensemble Jailbreak on Large Language Models5
A novel EnJa attack to hide harmful instructions using prompt-level jailbreak, boost the attack success rate using a gradient-based attack, and connect the two types of jailbreak attacks via a template-based connector are proposed.
- Toward Evaluating Robustness of Reinforcement Learning with Adversarial Policy5
This paper formulate four types of adversarial intrinsic regularizers—maximizing the adversarial state coverage, policy coverage, risk, or divergence—to discover potential vulnerabilities of the victim policy in a principled way and present a novel bias-reduction method to balance the extrinsic objective and the adversarial intrinsic regularizers adaptively.
- 5
- BrokenVideos: A Benchmark Dataset for Fine-Grained Artifact Localization in AI-Generated Videos4
It is demonstrated that training existing video artifact detection models and multi-modal large language models (MLLMs) on BrokenVideos substantially enhances their ability to localize corrupted regions within generated content.
- Simulated Ensemble Attack: Transferring Jailbreaks Across Fine-tuned Vision-Language Models4
The Simulated Ensemble Attack (SEA), a grey-box jailbreak framework that assumes full access to the base VLM but no knowledge of the fine-tuned target, is proposed, revealing that fine-tuning primarily induces localized parameter shifts around the base model, explaining why attacks optimized over a simulated neighborhood transfer effectively.
- T2UE: Generating Unlearnable Examples from Text Descriptions4
Text-to-Unlearnable Example (T2UE), a novel framework that enables users to generate UEs using only text descriptions, demonstrates the feasibility of ''zero-contact data protection', where personal data can be safeguarded based solely on their textual descriptions, eliminating the need for direct data exposure.
- 4
- 4
- Microwave Link Failures Prediction via LSTM-based Feature Fusion Network4
This work proposes a Long Short-Term Memory-based feature fusion network (LSTM-FFN) to fuse and encode both homophy and structural equivalence relationships in the LSTM temporal feature learning network to detect any link failures.
- Relationships Between Local Intrinsic Dimensionality and Tail Entropy4
This paper shows how LID is asymptotically related to the entropy of the lower tail of the distribution of distances from the query, and establishes tight relationships for cumulative Shannon entropy, entropy power, and their generalized Tsallis entropy variants.
- PaperAsk: A Benchmark for Reliability Evaluation of LLMs in Paper Search and Reading3
PaperAsk is introduced, a benchmark that systematically evaluates LLMs across four key research tasks: citation retrieval, content extraction, paper discovery, and claim verification, and GPT-4o, GPT-5, and Gemini-2.5-Flash under realistic usage conditions.
- NAP-Tuning: Neural Augmented Prompt Tuning for Adversarially Robust Vision-Language Models3
This work highlights the efficacy of internal feature-level intervention in prompt tuning for adversarial robustness, moving beyond input-side alignment approaches to create an adaptive defense mechanism that can identify and rectify adversarial perturbations across embedding spaces.
- From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving3
SERA is proposed, an LLM-powered framework that enables autonomous driving systems to self-evolve by repairing failure cases through targeted scenario recommendation, and consistently improves key metrics across multiple autonomous driving baselines.
- Towards Million-Scale Adversarial Robustness Evaluation With Stronger Individual Attacks3
This work proposes a novel individual attack method, Probability Margin Attack (PMA), which defines the adversarial margin in the probability space rather than the logits space, and analyzes the relationship between PMA and existing cross-entropy or logits-margin-based attacks, showing that PMA can outperform the current state-of-the-art individual methods.
- 3
- 3
- From Order to Distribution: An Exact Operator Framework for Forgetting in Continual Learning2
This work derives an exact operator identity for the forgetting quantity, establishing an unconditional upper bound, identifying the leading asymptotic term, and characterize the convergence rate up to constants, which relates to geometric properties of the task distribution.
- AudioMosaic: Contrastive Masked Audio Representation Learning2
Extensive experiments show that AudioMosaic achieves state-of-the-art performance on several standard audio benchmarks under both linear probing and fine-tuning and that integrating the pretrained AudioMosaic encoder into audio-language models improves performance on audio-language tasks.
- 2
- 2
- 2
- 2
- Expose Before You Defend: Unifying and Enhancing Backdoor Defenses via Exposed Models2
The results demonstrate the importance of backdoor exposure for backdoor defense, showing that the exposed models can significantly benefit a range of downstream defense tasks, including backdoor label detection, backdoor trigger recovery, backdoor model detection, and backdoor removal.
- End-to-End Anti-Backdoor Learning on Images and Time Series2
This paper builds upon Anti-Backdoor Learning (ABL) and introduces an innovative method, End-to-End Anti-Backdoor Learning (E2ABL), for robust training against backdoor attacks, which significantly improves on existing defenses and is effective against a broad range of backdoor attacks in both image and time series domains.
- CIM: Constrained Intrinsic Motivation for Sparse-Reward Continuous Control2
Constrained Intrinsic Motivation (CIM) is proposed to leverage readily attainable task priors to construct a constrained intrinsic objective, and at the same time, exploit the Lagrangian method to adaptively balance the intrinsic and extrinsic objectives via a simultaneous-maximization framework.
- Transfer of Automated Performance Feedback Models to Different Specimens in Virtual Reality Temporal Bone Surgery2
A simple method of feedback transfer between specimens in virtual reality temporal bone surgery is discussed and the accuracy and effectiveness of the transfer is validated through a user study.
- 2
- Extracting Real-time Feedback with Neural Networks for Simulation-based Learning.2
A neural network based method (NNFB) is proposed to extract feedback based on the adversarial technique that utilizes a bounded update to minimize a L1 regularized loss via back-propagation and empirically shows that NNFB can be used to extract simple, yet effective feedback.
- 2
- 2
- 2
- 1
- RedRFT: A Light-Weight Benchmark for Reinforcement Fine-Tuning-Based Red Teaming1
RedRFT, a lightweight benchmark designed to simplify and standardize the implementation and evaluation of RFT-based red teaming, is introduced, offering high-quality single-file red teaming implementations and modular PPO core components, such as the General Advantage Estimator.
- Optimizing Cross-Client Domain Coverage for Federated Instruction Tuning of Large Language Models1
This work empirically establishes that cross-client domain coverage, rather than data heterogeneity, is the pivotal factor in FedDIT dynamics and introduces FedDCA, an algorithm that explicitly maximizes this coverage through diversity-oriented client center selection and retrieval-based augmentation, constructing diverse, non-redundant cross-client instruction sets.
- 1
- FedAPT: Federated Adversarial Prompt Tuning for Vision-Language Models1
This work introduces Federated Adversarial Prompt Tuning (FedAPT), a novel method designed to enhance the adversarial robustness of FPT, and proposes a class-aware prompt generator that generates visual prompts from text prompts.
- MMFair: Fair Learning via Min-Min Optimization1
This work proposes MMFair, an algorithm that optimizes perturbations through a min-min optimization approach, which helps mitigate the bias problem of linear head, while leveraging the more powerful feature-learning capabilities of MLPs, leading to more stable and accurate classification results.
- Downstream Transfer Attack: Adversarial Attacks on Downstream Models with Pre-trained Vision Transformers1
Downstream Transfer Attack (DTA) is proposed, a novel attack method that leverages a pre-trained ViT model to craft the adversarial example and then applies the adversarial example to attack a fine-tuned version of the model on a downstream dataset.
- 1
- 1
- 1
- 1
- –
- Perturbation Effects on Robustness and Individual Fairness–
This work introduces RIFair, a black-box adversarial framework that leverages a decoupled perturbation strategy to construct semantically preserved yet unrobust and/or unfair instance pairs and formalizes Robust Individual Fairness (RIF), a necessary criterion for trustworthy model assessment.
- PixelSmile: Toward Fine-Grained Facial Expression Editing–
PixelSmile is proposed, a diffusion framework that disentangles expression semantics via fully symmetric joint training that achieves superior disentanglement and robust identity preservation, confirming its effectiveness for continuous, controllable, and fine-grained expression editing, while naturally supporting smooth expression blending.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- FedEGG: Federated Learning with Explicit Global Guidance–
A novel FL algorithm FedWalk is proposed that leverages an ex-ternal easy-to-converge task at the server side as a leash task to guide the local training of the clients.
- –
- –
- –
- –
- –
- –
- Machine learning with adversarial perturbations and noisy labels–
This thesis investigates machine learning models including DNNs and RFs, and their learning behavior through the lens of adversarial perturbations, and explores the possibility of using adversarial techniques to generate feedback.
- Development of communication software for intelligent robot based on VB–
Taking the ET-18 robot as an example, the paper presents a designing method of communication for robot and personal computer using VB, introducing the whole process from communication theory, interface designing to software development.
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.