What it measures

Native Korean (35k questions), Japanese, Arabic (40 tasks), and Indonesian multi-discipline knowledge benchmarks.

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

Overall question-weighted accuracy on ArabicMMLU. This result does not cover KMMLU, JMMLU or IndoMMLU.

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

Overall question-weighted accuracy on ArabicMMLU. This result does not cover KMMLU, JMMLU or IndoMMLU.

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. ArabicMMLU: Assessing Massive Multitask Language Understanding in Arabic — published result table ↗

    Koto, Fajri, Li, Haonan, Shatnawi, Sara, Doughman, Jad, Sadallah, Abdelrahman Boda, Alraeesi, Aisha et al. · 2024

    All 13 authors

    Koto, Fajri, Li, Haonan, Shatnawi, Sara, Doughman, Jad, Sadallah, Abdelrahman Boda, Alraeesi, Aisha, Almubarak, Khalid, Alyafeai, Zaid, Sengupta, Neha, Shehata, Shady, Habash, Nizar, Nakov, Preslav, Baldwin, Timothy

Reviewed 2026-10-26. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.