Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
You can not select more than 25 topics
Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
wdz
8c309fc65a
|
2 years ago | |
|---|---|---|
| .. | ||
| readme.md | 2 years ago | |
readme.md
Statistical analysis for code clone detection on BigCloneBench
We add a statistical analysis with t-test and effect size computation(Cohen's d) on code clone detection task. We compare CodeBERT fine-tuned by our approach with original GraphCodeBERT.
Pipeline
We provides our implementation to do t-test and effect size computation on code clone task.
run
python t_test_main.py
Result
For code clone detection, the results show that the improvement of CodeBERT fine-tuned with our approach over GraphCodeBERT is significant with p<0.001 and Cohen's d = 0.498.
TODO: We will add more statistical analysis results on other tasks.
This repository contains code for paper "Bridging Pre-trained Models and Downstream Tasks for Source Code Understanding". (Accepted to the 44th International Conference on Software Engineering (ICSE 2022))
Python
Contributors (1)