CoE-GRPO-7B (Ours)
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
2026.03 | 12.24 | 49.88 | 93.95 | 65.16 | 6.62 | 11.09 | 8.29 | |
2026.03 | 10.23 | 43.46 | 32.07 | 36.91 | 6.96 | 5.08 | 5.87 | |
2026.03 | 9.84 | 45.57 | 54.74 | 49.74 | 4.94 | 5.14 | 5.04 | |
2026.03 | 9.64 | 47.79 | 90.07 | 62.45 | 5 | 9.21 | 6.48 | |
2026.03 | 9.1 | 33.8 | 33.66 | 33.73 | 3.01 | 3.02 | 3.01 | |
2026.03 | 8 | 48.94 | 41.38 | 44.84 | 5.32 | 4.25 | 4.72 | |
2026.03 | 5.8 | 48.77 | 45.77 | 47.22 | 3.37 | 3.23 | 3.3 | |
2026.03 | 5.73 | 44.83 | 56.54 | 50.01 | 2.21 | 3.21 | 2.62 | |
2026.03 | 5.67 | 44.5 | 41.95 | 43.19 | 3.13 | 2.95 | 3.04 | |
2026.03 | 5.12 | 31.79 | 36.43 | 33.95 | 2.4 | 3.27 | 2.77 |