Search Results for author:
Found 13 papers, 5 papers with code
Threat Behavior Textual Search by Attention Graph Isomorphism
As such, analysts often resort to text search techniques to identify existing malware reports based on the symptoms they observe, exploiting the fact that malware samples share a lot of similarity, especially those from the same origin.
FedPIT: Towards Privacy-preserving and Few-shot Federated Instruction Tuning
Extensive experiments on real-world medical data demonstrate the effectiveness of FedPIT in improving federated few-shot performance while preserving privacy and robustness against data heterogeneity.
CodeArt: Better Code Models by Attention Regularization When Symbols Are Lacking
Our pre-trained model can improve the SOTAs in these tasks from 53% to 64%, 49% to 60%, and 74% to 94%, respectively.
Rapid Optimization for Jailbreaking LLMs via Subconscious Exploitation and Echopraxia
Large Language Models (LLMs) have become prevalent across diverse sectors, transforming human life with their extraordinary reasoning and comprehension abilities.
MULTIVERSE: Exposing Large Language Model Alignment Problems in Diverse Worlds
Given the low cost of our method, we are able to conduct a large scale study regarding LLM alignment issues in different worlds.
Make Them Spill the Beans! Coercive Knowledge Extraction from (Production) LLMs
Instead, it exploits the fact that even when an LLM rejects a toxic request, a harmful response often hides deep in the output logits.
When Federated Learning Meets Pre-trained Language Models' Parameter-Efficient Tuning Methods
To facilitate the research of PETuning in FL, we also develop a federated tuning framework FedPETuning, which allows practitioners to exploit different PETuning methods under the FL training paradigm conveniently.
Backdoor Vulnerabilities in Normally Trained Deep Learning Models
We leverage 20 different types of injected backdoor attacks in the literature as the guidance and study their correspondences in normally trained models, which we call natural backdoor vulnerabilities.
Parameter-Efficient Conformers via Sharing Sparsely-Gated Experts for End-to-End Speech Recognition
Experimental results show that the proposed model achieves competitive performance with 1/3 of the parameters of the encoder, compared with the full-parameter model.
DECK: Model Hardening for Defending Pervasive Backdoors
As such, using the samples derived from our attack in adversarial training can harden a model against these backdoor vulnerabilities.
Do Deep Neural Networks Always Perform Better When Eating More Data?
Data has now become a shortcoming of deep learning.
Constrained Optimization with Dynamic Bound-scaling for Effective NLPBackdoor Defense
We develop a novel optimization method for NLPbackdoor inversion.
Seizure Detection using Least EEG Channels by Deep Convolutional Neural Network
We design the SeizNet, a Convolutional Neural Network for seizure detection.
