Pitfalls of Evaluating Language Models with Open Benchmarks (2507.00460v1)
Abstract: Open LLM benchmarks, such as HELM and BIG-bench, offer standardized, transparent protocols that facilitate the fair comparison, reproducibility, and iterative advancement of LLMs (LMs). However, their openness also introduces critical and underexplored pitfalls. This study exposes these weaknesses by systematically constructing ``cheating'' models -- smaller variants of BART, T5, and GPT-2 fine-tuned directly on public test sets -- which achieve top rankings on a prominent open, holistic benchmark (HELM) despite poor generalization and limited practical utility. Our findings underscore three key insights: \ca high leaderboard performance on open benchmarks may not always reflect real-world effectiveness; \cb private or dynamic benchmarks must complement open evaluations to safeguard integrity; and \cc a fundamental reevaluation of current benchmarking practices is essential to ensure robust and trustworthy LM assessments.