Long-context language understanding on LongBench 14-subtask configuration
12.4Average Performance (14 Tasks)Softmax
Evaluation results
| Method | Links | |
|---|---|---|
| SoftmaxCondition=Softmax, Model scale=1.3 B2026.05 | 12.4 | |
| InterdomainCondition=Interdomain, Model scale=1.3 B2026.05 | 12.29 | |
| S4D-onlyCondition=S4D-only, Model scale=1.3 B2026.05 | 8.57 |