Are We Done with MMLU?

TL;DR

This paper reveals extensive errors in the MMLU benchmark, introduces a framework for error detection, and creates a corrected subset called MMLU-Redux to improve the benchmark's reliability.

Contribution

It identifies widespread errors in MMLU, develops a novel error annotation protocol, and provides a corrected dataset to enhance future benchmark assessments.

Findings

57% of Virology questions contain errors

6.49% of MMLU questions are erroneous

Re-annotated MMLU-Redux shows discrepancies with original performance metrics

Abstract

Maybe not. We identify and analyse errors in the popular Massive Multitask Language Understanding (MMLU) benchmark. Even though MMLU is widely adopted, our analysis demonstrates numerous ground truth errors that obscure the true capabilities of LLMs. For example, we find that 57% of the analysed questions in the Virology subset contain errors. To address this issue, we introduce a comprehensive framework for identifying dataset errors using a novel error annotation protocol. Then, we create MMLU-Redux, which is a subset of 5,700 manually re-annotated questions across all 57 MMLU subjects. We estimate that 6.49% of MMLU questions contain errors. Using MMLU-Redux, we demonstrate significant discrepancies with the model performance metrics that were originally reported. Our results strongly advocate for revising MMLU's error-ridden questions to enhance its future utility and reliability as a benchmark. https://huggingface.co/datasets/edinburgh-dawg/mmlu-redux-2.0.