Long-context understanding on Average MuSiQue, 2WikiMQA, MFQA-En, NarrativeQA, En.QA
42.48SubEM ScoreLogo-PO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Logo-POBackbone=Qwen2.5-14B-Instruct2025.08 | 42.48 | 42.46 | |
| LongMabBackbone=Qwen2.5-14B-Instruct2025.08 | 42.15 | 45.84 | |
| LongFaith-POBackbone=Qwen2.5-14B-Instruct2025.08 | 40.89 | 43.88 | |
| Vanilla LLMBackbone=Qwen2.5-14B-Instruct2025.08 | 40.64 | 39.57 | |
| SeaLong-POBackbone=Qwen2.5-14B-Instruct2025.08 | 40.33 | 37.88 | |
| LongReward-POBackbone=Qwen2.5-14B-Instruct2025.08 | 40.26 | 39.98 | |
| LongMab-SFTBackbone=Qwen2.5-14B-Instruct2025.08 | 38.75 | 41.81 | |
| LongAlpacaBackbone=Qwen2.5-14B-Instruct2025.08 | 38.12 | 39.98 | |
| LongAlignBackbone=Qwen2.5-14B-Instruct2025.08 | 35.27 | 40.4 |