Scholarlink

<p>Pre-trained code models lead the era of code intelligence, with multiple models designed with impressive performance. However, one important problem, data augmentation for code data that automatically helps developers prepare training data lacks study in this field. In this paper, we introduce a generic data augmentation framework, GenCode, to enhance the training of code understanding models. Simply speaking, GenCode follows a generation-and-selection paradigm to prepare useful training code data. Specifically, it employs code augmentation techniques to generate new code candidates first and then identifies important ones as the training data by influence scores. To evaluate the effectiveness of GenCode, we conduct experiments on four code understanding tasks&#xa0;(e.g., code clone detection) and three pre-trained code models&#xa0;(e.g., CodeT5) and two recent released code-specific Large Language Models (LLMs)&#xa0;(e.g., Qwen2.5-Coder). Compared to the <i>state-of-the-art (SOTA)</i> code augmentation method MixCode, GenCode produces pre-trained code models with 2.92% higher accuracy and 4.90% adversarial robustness on average. For code-specific LLMs, GenCode achieves an average improvement of 0.93% in accuracy and 0.98% in natural robustness.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GenCode: A generic data augmentation framework for boosting deep learning-based code understanding

  • 期刊论文

  • 发表日期：

    2026年2月10日

查看全文
  • Zeming Dong,
  • Qiang Hu,
  • Xiaofei Xie,
  • Maxime Cordy,
  • Mike Papadakis,
  • Yves Le Traon,
  • Jianjun Zhao

摘要

Pre-trained code models lead the era of code intelligence, with multiple models designed with impressive performance. However, one important problem, data augmentation for code data that automatically helps developers prepare training data lacks study in this field. In this paper, we introduce a generic data augmentation framework, GenCode, to enhance the training of code understanding models. Simply speaking, GenCode follows a generation-and-selection paradigm to prepare useful training code data. Specifically, it employs code augmentation techniques to generate new code candidates first and then identifies important ones as the training data by influence scores. To evaluate the effectiveness of GenCode, we conduct experiments on four code understanding tasks (e.g., code clone detection) and three pre-trained code models (e.g., CodeT5) and two recent released code-specific Large Language Models (LLMs) (e.g., Qwen2.5-Coder). Compared to the state-of-the-art (SOTA) code augmentation method MixCode, GenCode produces pre-trained code models with 2.92% higher accuracy and 4.90% adversarial robustness on average. For code-specific LLMs, GenCode achieves an average improvement of 0.93% in accuracy and 0.98% in natural robustness.