Holistic Long-Context Understanding on HELMET Holistic understanding
46.51HELMET Holistic Understanding (64K Context)PolicyLong
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PolicyLongEvaluation Stage=Pre-SFT, Base Model=Qwen2.5-3B2026.04 | 46.51 | 42.57 | 44.54 | |
| NExtLongEvaluation Stage=Pre-SFT, Base Model=Qwen2.5-3B2026.04 | 45.35 | 41.24 | 43.3 | |
| EntropyLongEvaluation Stage=Pre-SFT, Base Model=Qwen2.5-3B2026.04 | 43.88 | 40.08 | 41.98 | |
| Qwen2.5-3BEvaluation Stage=Pre-SFT, Base Model=Qwen2.5-3B2026.04 | 40.53 | 38.99 | 39.76 |