#ModelAccPaper
1LLaDA2.1-Mini SingleBD (Native)93.03link
2LLaDA2.1-Mini MultiBD (training-free)92.27link
3LLaDA2-Mini-CAP SingleBD (Native)91.74link
4LLaDA2-Mini-CAP MultiBD (training-free)91.21link
5HRM-Text: Efficient Pretraining Beyond Scaling84.50link
6Base – (original)45.94link
7Base OPM + XKD z (ours)44.70link
8Base OPM + XKD h (ours)44.12link
9Chat OPM + XKD z (ours)41.85link
10Chat – (original)41.32link
11Chat OPM + XKD h (ours)40.41link
12Base OPM + UT36.58link
13Chat OPM + UT33.59link
14Chat OPM3.26link
15Base OPM1.52link
GSM8K gsm8k Leaderboard