Next-token reasoning on OMNI-MATH Hard (val)
38.1AccuracyLoopRPT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LoopRPTModel=Ouro-2.6B, Reasoning Mode=Peak2026.03 | 38.1 | 4 | |
| LoopRPTModel=Ouro-2.6B, Reasoning Mode=Adap.2026.03 | 37.24 | 2.28 | |
| LoopRPTModel=Ouro-1.4B, Reasoning Mode=Adap.2026.03 | 34.82 | 3.07 | |
| LoopRPTModel=Ouro-1.4B, Reasoning Mode=Peak2026.03 | 34.74 | 4 | |
| PeakModel=Ouro-2.6B2026.03 | 34.52 | 4 | |
| Adap.Model=Ouro-2.6B2026.03 | 34.35 | 3.51 | |
| Adap.Model=Ouro-1.4B2026.03 | 33.91 | 3.75 | |
| PeakModel=Ouro-1.4B2026.03 | 33.79 | 4 | |
| VanillaModel=Qwen3-1.7B2026.03 | 19.19 | — | |
| +CoTModel=Qwen3-1.7B2026.03 | 7.44 | — |