In-Context Retrieval on RULER
99.55MQ ScoreFull
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| FullBackbone=GLM-4.7-Flash2026.04 | 99.55 | 99.75 | 53.6 | 66.25 | 100 | 95.2 | 85.73 | |
| AsyncTLSBackbone=GLM-4.7-Flash, Token Budget=5122026.04 | 95.1 | 98.8 | 48.2 | 59.5 | 99.88 | 87.53 | 81.5 | |
| DSBackbone=GLM-4.7-Flash, Token Budget=5122026.04 | 92.9 | 99.75 | 49 | 59.4 | 100 | 93.2 | 82.38 | |
| QuestBackbone=GLM-4.7-Flash, Token Budget=5122026.04 | 57.45 | 42.25 | 30.4 | 45.8 | 60.84 | 83.2 | 53.32 |