Long-context Understanding on LongBench (Average Score and Length Statistics)
6,711Avg Length (English Words)LongBench baseline
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LongBench baselineTest instances=4,750, Task Scope=21 tasks across 6 categories2026.03 | 6,711 | — | 13,386 | |
| LAVA2026.03 | — | 36.74 | — |