A5056795788- Software Testing and Debugging Techniques
- Software Engineering Research
- Advanced Malware Detection Techniques
- Adversarial Robustness in Machine Learning
- Artificial Intelligence in Games
- Software Reliability and Analysis Research
- Topic Modeling
Tianjin University
2021-2024
Hong Kong University of Science and Technology
2024
Guangzhou HKUST Fok Ying Tung Research Institute
2024
University of Hong Kong
2024
There are increasing uses of deep learning (DL) compilers to generate optimized code, boosting the runtime performance DL models on specific hardware. Like their traditional counterparts, can incorrect resulting in unexpected model behaviors that may cause catastrophic consequences mission-critical systems. On other hand, processed by differ fundamentally from imperative programs program logic is implicit. As such, various characteristics bugs arising need be revisited context compilers.
DL frameworks are the basis of constructing all programs and models, thus their bugs could lead to unexpected behaviors any program or model relying on them. Such a wide effect demonstrates necessity importance guaranteeing frameworks’ quality. Understanding characteristics framework is fundamental step for this quality assurance task, facilitating designing effective bug detection debugging approaches. Hence, in work, we conduct most large-scale study 1,000 from four popular diverse (i.e.,...
Question answering (QA) software uses information retrieval and natural language processing techniques to automatically answer questions posed by humans in a language. Like other AI-based software, QA may contain bugs. To test without human labeling, previous work extracts facts from question pairs generates new detect Nevertheless, the generated could be ambiguous, confusing, or with chaotic syntax, which are unanswerable for software. As result, relatively large proportion of reported bugs...
Deep Learning (DL) compilers are widely adopted to optimize advanced DL models for efficient deployment on diverse hardware. Their quality has a profound effect the of compiled models. A recent bug study shows that optimization high-level intermediate representations (IRs) is most error-prone compilation stage and bugs in this account 44.92% whole collected ones. However, existing testing techniques do not consider features related (e.g., IR), therefore weak exposing at stage. To bridge gap,...
Solidity compiler plays a key role in enabling the development of smart contract applications on Ethereum by governing syntax domain-specific language called and performing compilation optimization code.The correctness is critical fostering transparency, efficiency, trust industries reliant contracts.However, like other software systems, prone to bugs, which may produce incorrect bytecodes blockchain platforms, resulting severe security concerns.As for contracts, differs from compilers many...
Deep Learning (DL) compilers typically load a DL model and optimize it with intermediate representation.Existing compiler testing techniques mainly focus on optimization stages, but rarely explore bug detection at the loading stage. Effectively stage requires covering diverse usages of each operator from various libraries, which shares common objective library testing, indicating that embedded knowledge in tests is beneficial for compilers. In this work, we propose OPERA to extract such...