NLP Evaluation in trouble: On the Need to Measure LLM Data Contamination for each Benchmark

Sainz, Oscar, Campos, Jon Ander, García-Ferrero, Iker · arXiv · 2023