Mathematical Reasoning on GSM8K, MATH500, ASDiv, SVAMP, AMC23 (test)
81.96GSM8K AccuracyOuro-1.4B-STARS
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Ouro-1.4B-STARSRecurrents=42026.05 | 81.96 | 67.4 | 84.73 | 84.33 | 52.5 | 74.18 | |
| Qwen3-4BRecurrents=-2026.05 | 80.36 | 63 | 86.59 | 89 | 12.5 | 66.29 | |
| Ouro-1.4B-SFTRecurrents=42026.05 | 80.06 | 64.6 | 83.47 | 76.67 | 47.5 | 70.46 | |
| Ouro-1.4BRecurrents=22026.05 | 76.88 | 54.8 | 74.27 | 81.67 | 35 | 64.52 | |
| Ouro-1.4B-SFTRecurrents=22026.05 | 76.04 | 55.8 | 82.08 | 82.33 | 30 | 65.25 | |
| Ouro-1.4B-STARSRecurrents=22026.05 | 75.89 | 56.4 | 82.56 | 79.67 | 40 | 66.9 | |
| Qwen3-1.7BRecurrents=-2026.05 | 75.82 | 65.8 | 85.64 | 89 | 17.5 | 66.75 | |
| Ouro-1.4BRecurrents=42026.05 | 75.21 | 59.6 | 76.57 | 75.67 | 50 | 67.41 | |
| Qwen2.5-3B-InstructRecurrents=-2026.05 | 74.91 | 68.2 | 76.18 | 87 | 32.5 | 67.76 | |
| Ouro-1.4B-STARSRecurrents=82026.05 | 74.45 | 54.8 | 82.52 | 81 | 35 | 65.55 | |
| Llama-3.2-3B-InstructRecurrents=-2026.05 | 71.11 | 42.4 | 82.04 | 79.67 | 15 | 58.04 | |
| Ouro-1.4B-SFTRecurrents=82026.05 | 60.05 | 39.2 | 75.1 | 68 | 22.5 | 52.97 | |
| Ouro-1.4BRecurrents=82026.05 | 58.23 | 40.8 | 70.07 | 66.33 | 40 | 55.09 | |
| Gemma3-1b-itRecurrents=-2026.05 | 42.76 | 41.2 | 72.45 | 62.67 | 25 | 48.82 | |
| Recurrent-OLMo-2-0425Recurrents=82026.05 | 17.97 | 11.4 | 40 | 38.48 | 0 | 21.57 | |
| Recurrent-OLMo-2-0425Recurrents=22026.05 | 17.89 | 9 | 29.33 | 31.97 | 0 | 17.64 | |
| Recurrent-OLMo-2-0425Recurrents=42026.05 | 17.13 | 9.8 | 38.13 | 38.67 | 2.5 | 21.25 | |
| Recurrent-Llama-3.2Recurrents=22026.05 | 15.01 | 14.6 | 30.5 | 18.67 | 0 | 15.76 | |
| Recurrent-Llama-3.2Recurrents=42026.05 | 14.03 | 14.6 | 41.65 | 18.67 | 5 | 18.79 | |
| Recurrent-Llama-3.2Recurrents=82026.05 | 13.04 | 14 | 21.33 | 42.17 | 7.5 | 19.61 |