1 citations · 1 across the 3 of their papers we have counts for
7 papers
TokenTrace: Multi-Concept Attribution through Watermarked Token Recovery
Li Zhang, Shruti Agarwal, John Collomosse +2
Generative AI models pose a significant challenge to intellectual property (IP), as they can replicate unique artistic styles and concepts without attribution. While watermarking o…
BLO-Inst: Bi-Level Optimization Based Alignment of YOLO and SAM for Robust Instance Segmentation
Li Zhang, Pengtao Xie
The Segment Anything Model has revolutionized image segmentation with its zero-shot capabilities, yet its reliance on manual prompts hinders fully automated deployment. While integ…
SteganoBackdoor: Stealthy and Data-Efficient Backdoor Attacks on Language Models
Eric Xue, Ruiyi Zhang, Pengtao Xie
Modern language models remain vulnerable to backdoor attacks via poisoned data, where training inputs containing a trigger are paired with a target output, causing the model to rep…
BiDoRA: Bi-level Optimization-Based Weight-Decomposed Low-Rank Adaptation
Peijia Qin, Ruiyi Zhang, Pengtao Xie
Parameter-efficient fine-tuning (PEFT) is a flexible and efficient method for adapting large language models (LLMs) to downstream tasks. Among these methods, weight-decomposed low-…
Defense against Prompt Injection Attacks via Mixture of Encodings
Ruiyi Zhang, David Sullivan, Kyle Jackson +2
Large Language Models (LLMs) have emerged as a dominant approach for a wide range of NLP tasks, with their access to external information further enhancing their capabilities. Howe…
TapWeight: Reweighting Pretraining Objectives for Task-Adaptive Pretraining
Ruiyi Zhang, Sai Ashish Somayajula, Pengtao Xie
Large-scale general domain pretraining followed by downstream-specific finetuning has become a predominant paradigm in machine learning. However, discrepancies between the pretrain…