Logic Reasoning on Zebra riddles (test)
98.7AccuracyGlauber-UL2 (N=3)
Evaluation Results
| Method | Links | |
|---|---|---|
| Glauber-UL2 (N=3)Params=21M2026.05 | 98.7 | |
| MDLM (Top-K Prob.)Params=19M2026.05 | 98.5 | |
| MDLM (Top-K Margin)Params=19M2026.05 | 98.3 | |
| Glauber-UL2 (N=1)Params=21M2026.05 | 94.1 | |
| AR (w/ ordering)Params=42M2026.05 | 91.2 | |
| AR (w/o ordering)Params=42M2026.05 | 80.3 | |
| MDLM (vanilla)Params=19M2026.05 | 76.9 |