tianzhaotju

LEAM

We propose a novel DL-based mutation technique (LEAM), which adapts the syntax-guided encoder-decoder architecture to build two sub-models based on AST information, for better learning to represent real faults and ensuring syntactic correctness of mutation faults.

CODA

We propose a novel adversarial example generation technique (i.e., CODA) for testing deep code models. Its key idea is to use code differences between the target input and reference inputs to guide the generation of adversarial examples.

Deeplive

We propose Deeplive, which maximizes QoE through deep reinforcement learning, so it does not rely on fixed rules. To accelerate the training process of Deeplive, we further propose optimization including window completion with historical data and quick-start with rate-based algorithm.

CodeDenoise

In this work, we propose the first input denoising technique (i.e., CodeDenoise) for deep code models. Its key idea is to localize noisy identifiers in (likely) mispredicted inputs, and denoise such inputs by cleansing the located identifiers.

EMD

CoDefense

tianzhaotju.github.io