| Benchmark | Model | Rank | Results |
|---|---|---|---|
| arithmetic-reasoning-on-gsm8k | TIE (LLaDA + Dream; Entropy) | – | Accuracy: 83.62 |
| code-generation-on-humaneval | TIE (DreamCoder + DiffuCoder; Token change count) | – | Pass@1: 72.56 |
| common-sense-reasoning-on-winogrande | TIE (LLaDA + Dream; Top-1 probability) | – | Accuracy: 73.95 |
| mathematical-reasoning-on-math-500 | TIE (LLaDA + Dream; Token change count) | – | Accuracy: 48.6 |