#ModelAccuracyPaper
1Base – (original)80.98link
2Base OPM + XKD z (ours)80.40link
3Base OPM + XKD h (ours)80.00link
4Chat – (original)79.03link
5Chat OPM + XKD z (ours)78.96link
6Base OPM + UT78.35link
7Chat OPM + XKD h (ours)78.31link
8Chat OPM + UT76.73link
9Surefire-3B67.00link
10Panda-3B66.90link
11LLaMA-3.2-3B66.70link
12Panda-1B58.40link
13LLaMA-3.2-1B56.90link
14Surefire-1B56.60link
15Chat OPM34.25link
16Base OPM30.50link
HellaSwag hellaswag Leaderboard