KMMAU - Role/Profession: leaderboard

Metric: Accuracy (%; four-choice question on a speaker's role or profession; native Korean audio from KSS, KMSAV and the Seoul Corpus; single run at temperature 0.7). Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.

Top models

#ModelScore
1MiniCPM-o-4.578

Interactive version: theaggregate.ai/benchmark?slug=kmmau-role-profession · How It Works · Data refreshed daily, snapshot 2026-09-26.