What it measures

5,700 manually re-annotated MMLU questions fixing the ~6.5% ground-truth label errors present in original MMLU.

Reading the measurement

Accuracy is only part of the picture. Inspect unsupported answers, abstentions, language coverage and the evidence available to the model. A confident response can still be wrong, and a broad average can hide a specialist gap.

Published evaluation

Accuracy restricted to questions annotated correct in the named audited subject. The table compares original and corrected-instance scores; only the value after the arrow is imported. This is a subject-level audit, not overall accuracy across all 57 subjects and not the paper’s separate error-detection F1 task.

Where this fits in an evaluation suite

Read accuracy together with error and abstention rates. Domain averages can hide language gaps, specialist failures, and poorly calibrated confidence.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Accuracy restricted to questions annotated correct in the named audited subject. The table compares original and corrected-instance scores; only the value after the arrow is imported. This is a subject-level audit, not overall accuracy across all 57 subjects and not the paper’s separate error-detection F1 task.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. Are We Done with MMLU? — published result table ↗

    Gema, Aryo Pradipta, Leang, Joshua Ong Jun, Hong, Giwon, Devoto, Alessio, Mancino, Alberto Carlo Maria, Saxena, Rohit et al. · 2024

    All 16 authors

    Gema, Aryo Pradipta, Leang, Joshua Ong Jun, Hong, Giwon, Devoto, Alessio, Mancino, Alberto Carlo Maria, Saxena, Rohit, He, Xuanli, Zhao, Yu, Du, Xiaotang, Madani, Mohammad Reza Ghasemi, Barale, Claire, McHardy, Robert, Harris, Joshua, Kaddour, Jean, van Krieken, Emile, Minervini, Pasquale

Reviewed 2026-10-27. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.