Position: LLM Benchmark Datasets should be Contamination-Resistant

摘要

Benchmark datasets are critical for reproducible, reliable and discriminative evaluation of LLMs. However, recent studies reveal that many benchmark datasets are included in pretraining corpora, i.e. *contaminated*, which diminishes their value as a reliable measure of model generalization. In this position paper, we argue that benchmark datasets should be *contamination-resistant*, i.e. *unlearnable* but support *inference*. To accomplish this, we first underline the wide prevalence of benchmark dataset contamination and outline the properties of contamination-resistant datasets. Second, we highlight how the asymmetry between the inference and training pipelines in the Transformer architecture can be leveraged to support contamination-resistance. Third, we outline mathematical advancements to make these datasets interoperable across various LLM architectures. Based on the above, we call on the community to ensure the reliability of LLM benchmarking by: (i) advancing novel contamination-resistant methodologies, (ii) develop supporting methods and platforms, and (iii) adopt contamination-resistant benchmarks into existing evaluation pipelines.