Shiqing Ma
YOU?
Author Swipe
View article: Uncertainty Quantification for Multiple-Choice Questions is Just One-Token Deep
Uncertainty Quantification for Multiple-Choice Questions is Just One-Token Deep Open
View article: Photosensitive small intestinal submucosal hydrogels loaded with the KR-12-a5 peptide promote periodontal osteogenesis and antimicrobial activity
Photosensitive small intestinal submucosal hydrogels loaded with the KR-12-a5 peptide promote periodontal osteogenesis and antimicrobial activity Open
View article: Rethinking Technology Stack Selection with AI Coding Proficiency
Rethinking Technology Stack Selection with AI Coding Proficiency Open
Large language models (LLMs) are now an integral part of software development workflows and are reshaping the whole process. Traditional technology stack selection has not caught up. Most of the existing selection methods focus solely on t…
View article: DeCoMa: Detecting and Purifying Code Dataset Watermarks through Dual Channel Code Abstraction
DeCoMa: Detecting and Purifying Code Dataset Watermarks through Dual Channel Code Abstraction Open
Watermarking is a technique to help identify the source of data points, which can be used to help prevent the misuse of protected datasets. Existing methods on code watermarking, leveraging the idea from the backdoor research, embed stealt…
View article: The Foundation Cracks: A Comprehensive Study on Bugs and Testing Practices in LLM Libraries
The Foundation Cracks: A Comprehensive Study on Bugs and Testing Practices in LLM Libraries Open
Large Language Model (LLM) libraries have emerged as the foundational infrastructure powering today's AI revolution, serving as the backbone for LLM deployment, inference optimization, fine-tuning, and production serving across diverse app…
View article: EDITOR: Effective and Interpretable Prompt Inversion for Text-to-Image Diffusion Models
EDITOR: Effective and Interpretable Prompt Inversion for Text-to-Image Diffusion Models Open
Text-to-image generation models~(e.g., Stable Diffusion) have achieved significant advancements, enabling the creation of high-quality and realistic images based on textual descriptions. Prompt inversion, the task of identifying the textua…
View article: VIDSTAMP: A Temporally-Aware Watermark for Ownership and Integrity in Video Diffusion Models
VIDSTAMP: A Temporally-Aware Watermark for Ownership and Integrity in Video Diffusion Models Open
Video diffusion models can generate realistic and temporally consistent videos. This raises concerns about provenance, ownership, and integrity. Watermarking can help address these issues by embedding metadata directly into the content. To…
View article: Exposing Product Bias in LLM Investment Recommendation
Exposing Product Bias in LLM Investment Recommendation Open
Large language models (LLMs), as a new generation of recommendation engines, possess powerful summarization and data analysis capabilities, surpassing traditional recommendation systems in both scope and performance. One promising applicat…
View article: Holistic Audit Dataset Generation for LLM Unlearning via Knowledge Graph Traversal and Redundancy Removal
Holistic Audit Dataset Generation for LLM Unlearning via Knowledge Graph Traversal and Redundancy Removal Open
In recent years, Large Language Models (LLMs) have faced increasing demands to selectively remove sensitive information, protect privacy, and comply with copyright regulations through unlearning, by Machine Unlearning. While evaluating unl…
View article: The Invisible Hand: Unveiling Provider Bias in Large Language Models for Code Generation
The Invisible Hand: Unveiling Provider Bias in Large Language Models for Code Generation Open
Large Language Models (LLMs) have emerged as the new recommendation engines, surpassing traditional methods in both capability and scope, particularly in code generation. In this paper, we reveal a novel provider bias in LLMs: without expl…
View article: An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer
An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer Open
View article: The Invisible Hand: Unveiling Provider Bias in Large Language Models for Code Generation
The Invisible Hand: Unveiling Provider Bias in Large Language Models for Code Generation Open
View article: MLLM-as-a-Judge for Image Safety without Human Labeling
MLLM-as-a-Judge for Image Safety without Human Labeling Open
Image content safety has become a significant challenge with the rise of visual media on online platforms. Meanwhile, in the age of AI-generated content (AIGC), many image generation models are capable of producing harmful content, such as…
View article: Continuous Concepts Removal in Text-to-image Diffusion Models
Continuous Concepts Removal in Text-to-image Diffusion Models Open
Text-to-image diffusion models have shown an impressive ability to generate high-quality images from input textual descriptions. However, concerns have been raised about the potential for these models to create content that infringes on co…
View article: DREAM: Debugging and Repairing AutoML Pipelines
DREAM: Debugging and Repairing AutoML Pipelines Open
Deep Learning models have become an integrated component of modern software systems. In response to the challenge of model design, researchers proposed Automated Machine Learning (AutoML) systems, which automatically search for model archi…
View article: Speculative Coreset Selection for Task-Specific Fine-tuning
Speculative Coreset Selection for Task-Specific Fine-tuning Open
Task-specific fine-tuning is essential for the deployment of large language models (LLMs), but it requires significant computational resources and time. Existing solutions have proposed coreset selection methods to improve data efficiency …
View article: An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer
An Optimizable Suffix Is Worth A Thousand Templates: Efficient Black-box Jailbreaking without Affirmative Phrases via LLM as Optimizer Open
Despite prior safety alignment efforts, mainstream LLMs can still generate harmful and unethical content when subjected to jailbreaking attacks. Existing jailbreaking methods fall into two main categories: template-based and optimization-b…
View article: UNIT: Backdoor Mitigation via Automated Neural Distribution Tightening
UNIT: Backdoor Mitigation via Automated Neural Distribution Tightening Open
Deep neural networks (DNNs) have demonstrated effectiveness in various fields. However, DNNs are vulnerable to backdoor attacks, which inject a unique pattern, called trigger, into the input to cause misclassification to an attack-chosen t…
View article: COSTELLO: Contrastive Testing for Embedding-Based Large Language Model as a Service Embeddings
COSTELLO: Contrastive Testing for Embedding-Based Large Language Model as a Service Embeddings Open
Large language models have gained significant popularity and are often provided as a service (i.e., LLMaaS). Companies like OpenAI and Google provide online APIs of LLMs to allow downstream users to create innovative applications. Despite …
View article: Efficient DNN-Powered Software with Fair Sparse Models
Efficient DNN-Powered Software with Fair Sparse Models Open
With the emergence of the Software 3.0 era, there is a growing trend of compressing and integrating large models into software systems, with significant societal implications. Regrettably, in numerous instances, model compression technique…
View article: CITADEL: Context Similarity Based Deep Learning Framework Bug Finding
CITADEL: Context Similarity Based Deep Learning Framework Bug Finding Open
With the application of deep learning technology, tools of DL framework testing are in high demand. Existing DL framework testing tools have limited coverage of bug types. For example, they lack the capability of effectively finding perfor…
View article: MeanSparse: Post-Training Robustness Enhancement Through Mean-Centered Feature Sparsification
MeanSparse: Post-Training Robustness Enhancement Through Mean-Centered Feature Sparsification Open
We present a simple yet effective method to improve the robustness of both Convolutional and attention-based Neural Networks against adversarial examples by post-processing an adversarially trained model. Our technique, MeanSparse, cascade…
View article: From Effectiveness to Efficiency: Uncovering Linguistic Bias in Large Language Model-based Code Generation
From Effectiveness to Efficiency: Uncovering Linguistic Bias in Large Language Model-based Code Generation Open
Large Language Models (LLMs) have demonstrated promising capabilities for code generation. While existing benchmarks evaluate the correctness and efficiency of LLM-generated code, the potential linguistic bias - where code quality varies b…
View article: Invisible Backdoor Attack against Self-supervised Learning
Invisible Backdoor Attack against Self-supervised Learning Open
Self-supervised learning (SSL) models are vulnerable to backdoor attacks. Existing backdoor attacks that are effective in SSL often involve noticeable triggers, like colored patches or visible noise, which are vulnerable to human inspectio…
View article: How to Trace Latent Generative Model Generated Images without Artificial Watermark?
How to Trace Latent Generative Model Generated Images without Artificial Watermark? Open
Latent generative models (e.g., Stable Diffusion) have become more and more popular, but concerns have arisen regarding potential misuse related to images generated by these models. It is, therefore, necessary to analyze the origin of imag…
View article: Merlin: Multi-tier Optimization of eBPF Code for Performance and Compactness
Merlin: Multi-tier Optimization of eBPF Code for Performance and Compactness Open
View article: LOTUS: Evasive and Resilient Backdoor Attacks through Sub-Partitioning
LOTUS: Evasive and Resilient Backdoor Attacks through Sub-Partitioning Open
Backdoor attack poses a significant security threat to Deep Learning applications. Existing attacks are often not evasive to established backdoor detection techniques. This susceptibility primarily stems from the fact that these attacks ty…
View article: Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift
Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift Open
Diffusion models (DM) have become state-of-the-art generative models because of their capability of generating high-quality images from noises without adversarial training. However, they are vulnerable to backdoor attacks as reported by re…
View article: Rapid Optimization for Jailbreaking LLMs via Subconscious Exploitation and Echopraxia
Rapid Optimization for Jailbreaking LLMs via Subconscious Exploitation and Echopraxia Open
Large Language Models (LLMs) have become prevalent across diverse sectors, transforming human life with their extraordinary reasoning and comprehension abilities. As they find increased use in sensitive tasks, safety concerns have gained w…
View article: Gradient Shaping: Enhancing Backdoor Attack Against Reverse Engineering
Gradient Shaping: Enhancing Backdoor Attack Against Reverse Engineering Open
Most existing methods to detect backdoored machine learning (ML) models take one of the two approaches: trigger inversion (aka.reverse engineer) and weight analysis (aka.model diagnosis).In particular, the gradient-based trigger inversion …