Data Contamination Can Cross Language Barriers

Feng Yao, Yufan Zhuang, Zihao Sun +3

The opacity in developing large language models (LLMs) is raising growing concerns about the potential contamination of public benchmarks in the pre-training data. Existing contami…