Long-context performance evaluation on LongBench
16GovReport Score
Hybrid Gated DeltaNet + M2RNN-3
Updated 4mo ago
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 16 | 18.1 | 14 | 43 | 41.5 | 24.7 | 12.7 | 13.9 | 23 | ||
2026.03 | 10.5 | 15.8 | 13.2 | 29 | 31.7 | 12.1 | 10.7 | 13.3 | 17.1 | |
2026.03 | 10 | 4.8 | 10.2 | 16 | 9.4 | 9.2 | 10.5 | 12 | 10.3 | |
2026.03 | 10 | 14 | 7.2 | 17.5 | 35.5 | 14.9 | 13.4 | 10.3 | 15.4 | |
2026.03 | 9.8 | 9 | 7.5 | 22.5 | 13.3 | 12.6 | 13.2 | 9.8 | 12.2 | |
| 9.5 | 14.8 | 17.6 | 47 | 32.1 | 19.4 | 7.3 | 8.5 | 19.5 | ||
2026.03 | 9.3 | 15.2 | 9.2 | 0.5 | 12.2 | 5 | 17.9 | 14.9 | 10.5 | |
2026.03 | 8.2 | 16.8 | 16.1 | 19.5 | 28.7 | 15.9 | 8.7 | 9.2 | 15.4 | |
2026.03 | 7.6 | 18.2 | 12 | 19 | 34.1 | 21.3 | 14.9 | 11.8 | 17.3 | |
2026.03 | 7 | 16.5 | 10.8 | 1 | 21.4 | 8.5 | 12.5 | 9 | 10.8 | |
2026.03 | 6.6 | 14.3 | 9.9 | 26.5 | 23.2 | 11.5 | 13.7 | 8.9 | 14.3 | |
2026.03 | 6.3 | 16.4 | 10.8 | 22 | 24.3 | 13.2 | 10.7 | 10.1 | 14.2 | |
2026.03 | 5.6 | 5.7 | 6.2 | 11 | 9.7 | 4.6 | 12 | 9.9 | 8.1 |