Variable Tracking on RULER 8k
79.28F1 ScoreRephrasing
Evaluation Results
| Method | Links | |
|---|---|---|
| RephrasingModel (Max Context)=Llama-3-8B (262k)2025.06 | 79.28 | |
| PCDModel (Max Context)=Llama-3-8B (262k)2025.06 | 77.92 | |
| DoLa-LowModel (Max Context)=Llama-3-8B (262k)2025.06 | 72.93 | |
| DoLa-HighModel (Max Context)=Llama-3-8B (262k)2025.06 | 72.87 | |
| BaseModel (Max Context)=Llama-3-8B (1048k)2025.06 | 72.78 | |
| Beam-SearchModel (Max Context)=Llama-3-8B (262k)2025.06 | 71.25 | |
| BaseModel (Max Context)=Llama-3-8B (262k)2025.06 | 71.21 | |
| PCDModel (Max Context)=Llama-3-8B (1048k)2025.06 | 71.19 | |
| MsPoEModel (Max Context)=Llama-3-8B (262k)2025.06 | 70.03 | |
| PCDModel (Max Context)=Llama-3-8B (8k)2025.06 | 60.07 | |
| BaseModel (Max Context)=Llama-3-8B (8k)2025.06 | 57.93 | |
| SegRModel (Max Context)=Llama-3-8B (262k)2025.06 | 0 |