#ModelAccuracyPaper
1Base OPM + XKD z (ours)69.40link
2Base – (original)69.39link
3Base OPM + XKD h (ours)68.36link
4Base OPM + UT67.57link
5Chat – (original)62.68link
6Chat OPM + XKD z (ours)62.15link
7Chat OPM + XKD h (ours)61.51link
8Chat OPM + UT60.94link
9HRM-Text: Efficient Pretraining Beyond Scaling60.70link
10Chat OPM31.17link
11Base OPM26.98link
MMLU mmlu Leaderboard