Towards More Complete Constraints for Deep Learning Library Testing via Complementary Set Guided Refinement
This program is tentative and subject to change.
Deep learning libraries such as PyTorch and TensorFlow play a crucial role in modern artificial intelligence systems. Recently, various testing techniques have been proposed to ensure their reliability. They often model the inputs of tensor operations as a set of constraints to guide the generation of valid test cases. However, these constraints may overly narrow the search space, resulting in incomplete testing for DL operations. This paper introduces Complementary Set Guided Refinement, an approach to enhance the completeness of constraints for DL operators. Its basic idea is to identify the complementary set of a constraint, and observe if the complementary set can yield valid test cases. If such test cases are found, it indicates that the original constraint is incomplete and suggests a potential direction for refinement. Based on this idea, we design an automatic constraint refinement tool called DeepConstr, which adopts genetic algorithm to iteratively refine constraints for better completeness. We evaluated it on two widely-tested DL libraries, i.e., PyTorch and TensorFlow. In total, DeepConstr discovered 85 unknown bugs, out of which 69 were confirmed, with 45 fixed. Compared to state-of-the-art fuzzers, DeepConstr achieved increased coverage for 43.44% of operators supported by NNSmith, and 59.16% ofoperators supported by NeuRI.
This program is tentative and subject to change.
Wed 18 SepDisplayed time zone: Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna change
13:30 - 14:50 | |||
13:30 20mTalk | UPBEAT: Test Input Checks of Q# Quantum Libraries Technical Papers Tianmin Hu Northwest University, Guixin Ye Northwest University, Zhanyong Tang Northwest University, Shin Hwei Tan Concordia University, Huanting Wang University of Leeds, UK, Meng Li Hefei University of Technology, Zheng Wang University of Leeds DOI | ||
13:50 20mTalk | Towards More Complete Constraints for Deep Learning Library Testing via Complementary Set Guided Refinement Technical Papers Gwihwan Go Tsinghua University, Chijin Zhou Tsinghua University, Quan Zhang Tsinghua University, Xiazijian Zou Central South University, Heyuan Shi Central South University, Yu Jiang Tsinghua University | ||
14:10 20mTalk | AsFuzzer: Differential Testing of Assemblers with Error-Driven Grammar Inference Technical Papers Hyungseok Kim The Affiliated Institute of ETRI, Soomin Kim KAIST, Jungwoo Lee KAIST, Sang Kil Cha KAIST | ||
14:30 20mTalk | Ma11y: A Mutation Framework for Web Accessibility Testing Technical Papers Mahan Tafreshipour University of California, Irvine, Anmol Vilas Deshpande University of California at Irvine, Forough Mehralian University of California at Irvine, Iftekhar Ahmed University of California, Irvine, Sam Malek University of California at Irvine DOI | ||