12 papers
VLTest: Finding Joint Image-Text Failures in Vision-Language Models
Bin Duan, Jintao Lin, Guowei Yang
Vision-language models (VLMs) are increasingly embedded in software systems. Testing them raises two problems. Mutating either input can change the expected answer, so an output ch…
Towards Effective Black-Box Adversarial Attacks on Deep Code Models via Structural and Identifier Perturbations
Bin Duan, Jintao Lin, Dan Dongseong Kim +1
Deep code models (DCMs) are increasingly embedded in code intelligence tasks. However, their robustness under adversarial attacks remains insufficiently understood. Prior black-box…
Testing Deep Learning Library APIs via Cross-Framework Differential Fuzzing
Bin Duan, Ruican Dong, Naipeng Dong +2
Deep learning libraries underpin many safety- and reliability-critical applications, yet existing API-level testing techniques often rely on intra-library properties or CPU--GPU di…
Harnessing LLMs for Document-Guided Fuzzing of Python Libraries
Bin Duan, Tarek Mahmud, Meiru Che +4
Python libraries underpin deep learning, scientific computing, data analysis, and computer vision, making their reliability critical to downstream applications. Testing their APIs…
Toward a Generalized Defense Across Sparse, Continuous, and Structured Parameter Attacks
Bin Duan, Zeyu Bai, Guowei Yang
Deep neural networks are increasingly deployed across heterogeneous and partially untrusted environments, where models are distributed through cloud storage, CI/CD pipelines, conta…
Testing Neural Networks via Bayesian-Guided Exploration of Decision Landscapes
Bin Duan, Meiru Che, Guowei Yang
As neural networks are increasingly deployed in safety-critical domains, testing is essential to evaluate and improve their reliability. Existing testing methods, whether black-box…
Latent Anchor-Driven Test Generation for Deep Neural Networks
Bin Duan, Matthew B. Dwyer, Guowei Yang
Deep Neural Networks (DNNs) are increasingly being deployed in security-critical and safety-sensitive applications, which makes rigorous testing essential to identify and mitigate…
XAMT: Cross-Framework API Matching for Testing Deep Learning Libraries
Bin Duan, Ruican Dong, Naipeng Dong +2
Deep learning powers critical applications such as autonomous driving, healthcare, and finance, where the correctness of underlying libraries is essential. Bugs in widely used deep…
Harnessing LLMs for Document-Guided Fuzzing of OpenCV Library
Bin Duan, Tarek Mahmud, Meiru Che +4
The combination of computer vision and artificial intelligence is fundamentally transforming a broad spectrum of industries by enabling machines to interpret and act upon visual da…
Generating Realistic, Diverse, and Fault-Revealing Inputs with Latent Space Interpolation for Testing Deep Neural Networks
Bin Duan, Matthew B. Dwyer, Guowei Yang
Deep Neural Networks (DNNs) have been widely employed across various domains, including safety-critical systems, necessitating comprehensive testing to ensure their reliability. Al…
Enhancing LLM Code Generation with Ensembles: A Similarity-Based Selection Approach
Tarek Mahmud, Bin Duan, Corina Pasareanu +1
Ensemble learning has been widely used in machine learning to improve model robustness, accuracy, and generalization, but has not yet been applied to code generation tasks with lar…
Automated Update of Android Deprecated API Usages with Large Language Models
Tarek Mahmud, Bin Duan, Meiru Che +3
Android apps rely on application programming interfaces (APIs) to access various functionalities of Android devices. These APIs however are regularly updated to incorporate new fea…