English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Automated program repair (APR) aims to fix software bugs automaticallywithout human debugging efforts and plays a crucial role in softwaredevelopment and maintenance. Despite promising, APR is still challenged by along-standing overfitting problem (i.e., the generated patch is plausible butoverfitting). Various techniques have thus been proposed to address theoverfitting problem. Recently, researchers have employed BERT to extract codefeatures, which are then used to train a classifier for patch correctnessprediction. However, BERT is restricted to feature extraction for classifiertraining without benefiting from the training process, potentially generatingsub-optimal vector representations for patched code snippets. In this paper, wepropose APPT, a pre-trained model-based automated patch correctness assessmenttechnique by both pre-training and fine-tuning. APPT adopts a pre-trained modelas the encoder stack, followed by an LSTM stack and a deep learning classifier.More importantly, the pre-trained model is fine-tuned in conjunction with othercomponents as a whole pipeline to fully adapt it specifically for reasoningabout patch correctness. We conduct an extensive experiment on 1,183 Defects4Jpatches and the experimental results show that APPT achieves predictionaccuracy of 79.7% and recall of 83.2%, outperforming CACHE by 4.3% and 6.7%.Our additional investigation on 49,694 real-world patches shows that APPTachieves the optimum performance compared with existing representation learningtechniques. We further investigate the impact of each component and find thatthey all positively contribute to APPT, e.g., the fine-tuning process and theLSTM stack increase F1-score by 10.22% and 4.11%, respectively. We also provethat adopting advanced pre-trained models can further provide substantialadvancement, highlighting the generalizability of APPT.

APPT: Boosting Automated Patch Correctness Prediction via Fine-tuning Pre-trained Models