Long-context Understanding on Average Overall
40.95SubEMLongMab
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LongMabBackbone=Llama-3.1-8B-Instruct2025.08 | 40.95 | 47.43 | |
| LongMabBackbone=Qwen-2.5-7B-Instruct2025.08 | 37.06 | 42.26 | |
| SeaLong-POBackbone=Qwen-2.5-7B-Instruct2025.08 | 36.71 | 26.56 | |
| LongFaith-POBackbone=Qwen-2.5-7B-Instruct2025.08 | 34.85 | 34.19 | |
| LongFaith-POBackbone=Llama-3.1-8B-Instruct2025.08 | 34.79 | 43.15 | |
| LongReward-POBackbone=Qwen-2.5-7B-Instruct2025.08 | 34.51 | 33.69 | |
| Logo-POBackbone=Qwen-2.5-7B-Instruct2025.08 | 34.43 | 37.92 | |
| LongReward-POBackbone=Llama-3.1-8B-Instruct2025.08 | 34.34 | 39.89 | |
| Logo-POBackbone=Llama-3.1-8B-Instruct2025.08 | 34.13 | 42.69 | |
| LongAlignBackbone=Llama-3.1-8B-Instruct2025.08 | 33.99 | 39.34 | |
| SeaLong-POBackbone=Llama-3.1-8B-Instruct2025.08 | 33.58 | 42.04 | |
| Vanilla LLMBackbone=Llama-3.1-8B-Instruct2025.08 | 33.23 | 41.03 | |
| LongMab-SFTBackbone=Llama-3.1-8B-Instruct2025.08 | 32.95 | 40.48 | |
| LongAlpacaBackbone=Llama-3.1-8B-Instruct2025.08 | 32.8 | 40.03 | |
| Vanilla LLMBackbone=Qwen-2.5-7B-Instruct2025.08 | 32.03 | 33.29 | |
| LongAlpacaBackbone=Qwen-2.5-7B-Instruct2025.08 | 31.9 | 34.81 | |
| LongMab-SFTBackbone=Qwen-2.5-7B-Instruct2025.08 | 31.82 | 38.38 | |
| LongAlignBackbone=Qwen-2.5-7B-Instruct2025.08 | 28.88 | 36.09 |