Multi-objective Retrieval on NeedleBench M-RT (32k context stress slice)
100Slot-set Exactness
FullKV
Updated 2mo ago
Evaluation Results
| Method | Links | ||
|---|---|---|---|
2026.05 | 100 | 37.5 | |
2026.05 | 100 | 75 | |
2026.05 | 95.8 | 37.5 | |
2026.05 | 95.8 | 37.5 | |
2026.05 | 87.5 | 0 | |
2026.05 | 75 | 50 | |
2026.05 | 75 | 62.5 | |
2026.05 | 72.2 | 54.2 | |
2026.05 | 72.2 | 54.2 | |
2026.05 | 66.7 | 37.5 | |
2026.05 | 66.7 | 50 | |
2026.05 | 66.7 | 50 | |
2026.05 | 62.5 | 25 | |
2026.05 | 61.1 | 25 | |
2026.05 | 61.1 | 25 | |
2026.05 | 54.2 | 0 | |
2026.05 | 52.8 | 25 | |
2026.05 | 52.8 | 25 | |
2026.05 | 50 | 37.5 | |
2026.05 | 41.7 | 0 |