model generalization

The article addresses the challenges in evaluating large language models, including data contamination, sensitivity to prompts, and the high cost of benchmark creation. The authors propose a new approach that…