Are LLMs Better than Reported? Detecting Label Errors and Mitigating Their Effect on Model Performance

Omer Nahum, Nitay Calderon, Orgad Keller +2

NLP benchmarks rely on standardized datasets for training and evaluating models and are crucial for advancing the field. Traditionally, expert annotations ensure high-quality label…