|国家预印本平台
首页|Rethinking the effects of data contamination in Code Intelligence

Rethinking the effects of data contamination in Code Intelligence

Rethinking the effects of data contamination in Code Intelligence

来源:Arxiv_logoArxiv
英文摘要

In recent years, code intelligence has gained increasing importance in the field of automated software engineering. Meanwhile, the widespread adoption of Pretrained Language Models (PLMs) and Large Language Models (LLMs) has raised concerns regarding data contamination and its potential impact on model performance evaluation. This paper presents a systematic empirical study to investigate the fine-grained data contamination on code intelligence tasks. Our study involves diverse representative PLMs, namely RoBERTa and GPT-2, and LLMs, namely LLaMA and StarCoder, covering three major tasks: code translation, code generation, and code summarization. We categorize contamination scenarios into four types according to the code intelligence practice, namely input-only, output-only, unpaired, and paired contamination settings, and construct corresponding experimental and control groups for exploration. Experimental results show that, under the pre-training, fine-tuning, and inference paradigm adopted by PLMs, even deliberately injecting paired contamination does not lead to significant performance overestimation. But direct inference or small-scale fine-tuning uncovers the contamination effects. In contrast, LLMs with pre-training and inference paradigm are significantly affected by the paired contamination. Apart from the above, other contamination scenarios have no impact on both PLMs and LLMs. Our findings challenge the conventional belief that contamination inevitably leads to performance overestimation, providing new insights into the evaluation and deployment of code intelligence models.

Zhen Yang、Hongyi Lin、Yifan He、Jie Xu、Zeyu Sun、Shuo Liu、Pengpeng Wang、Zhongxing Yu、Qingyuan Liang

计算技术、计算机技术

Zhen Yang,Hongyi Lin,Yifan He,Jie Xu,Zeyu Sun,Shuo Liu,Pengpeng Wang,Zhongxing Yu,Qingyuan Liang.Rethinking the effects of data contamination in Code Intelligence[EB/OL].(2025-06-03)[2025-07-01].https://arxiv.org/abs/2506.02791.点此复制

评论