#ModelAccuracyPaper
1Base – (original)74.35link
2Base OPM + XKD z (ours)73.77link
3Base OPM + XKD h (ours)73.37link
4Chat OPM + XKD z (ours)73.01link
5Chat – (original)72.77link
6Base OPM + UT72.39link
7Chat OPM + XKD h (ours)72.22link
8Chat OPM + UT71.19link
9Panda-3B63.20link
10LLaMA-3.2-3B62.50link
11Surefire-3B60.50link
12Chat OPM60.38link
13Panda-1B58.60link
14Surefire-1B57.50link
15LLaMA-3.2-1B57.10link
16Base OPM56.75link
Winogrande winogrande Leaderboard