一句话：提出克隆感知对比学习方法 CONCORD，用于源代码自监督预训练，以提升克隆与缺陷检测等下游任务表示。
源代码深度学习模型近年来在软件工程任务中展现出巨大潜力。为了获得可迁移的通用代码表示，自监督预训练受到关注，并可用于克隆检测、缺陷检测等下游任务。本文提出 CONCORD——一种克隆感知的对比学习（clone-aware contrastive learning）方法，用于源代码表示学习。其核心思想是在自监督预训练中引入克隆感知的对比目标，使模型学习到对代码克隆关系敏感的表示。摘要未具体说明模型架构、预训练数据、评估基准及量化结果，仅指出该类表示对多种下游 SE 任务有价值。
Deep Learning (DL) models to analyze source code have shown immense promise during the past few years. More recently, self-supervised pre-training has gained traction for learning generic code representations valuable for many downstream SE tasks, such as clone and bug detection.
@inproceedings{paperbot8064,
title = {CONCORD: Clone-Aware Contrastive Learning for Source Code},
author = {Yangruibo Ding and Saikat Chakraborty and Luca Buratti and Saurabh Pujar and Alessandro Morari and Gail Kaiser and Baishakhi Ray},
booktitle = {ISSTA},
year = {2023},
doi = {10.1145/3597926.3598035}
}