CoE-GRPO-7B (Ours)
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
2026.03 | 18.75 | 48.68 | 93.54 | 64.03 | 7.7 | 13.8 | 9.88 | |
2026.03 | 11.44 | 43.18 | 59.02 | 49.87 | 5.72 | 6.27 | 5.98 | |
2026.03 | 10.62 | 34.94 | 35.42 | 35.18 | 3.27 | 3.34 | 3.3 | |
2026.03 | 10.42 | 47.35 | 91.35 | 63.49 | 5.8 | 9.72 | 7.26 | |
2026.03 | 9.57 | 39.54 | 35.1 | 37.19 | 5.54 | 3.86 | 4.55 | |
2026.03 | 7.95 | 49.24 | 41.57 | 45.08 | 5.38 | 4.41 | 4.85 | |
2026.03 | 7 | 43.64 | 37.94 | 40.59 | 3.6 | 3.07 | 3.31 | |
2026.03 | 6.77 | 44.26 | 59.33 | 50.7 | 3.76 | 4.64 | 4.15 | |
2026.03 | 6.27 | 44.96 | 41.64 | 43.24 | 3.57 | 3.3 | 3.43 | |
2026.03 | 5.3 | 31.46 | 37.04 | 34.02 | 2.83 | 3.16 | 2.99 |