2024
An Open-Source Data Contamination Report for Large Language Models
EMNLP 2024finding
Data contamination in model evaluation has become increasingly prevalent with the growing popularity of large language models. It allows models to “cheat” via memorisation instead of displaying true capabilities. Therefore, contamination analysis has become an crucial part of reliable model evaluati…