Long-context Question Answering on MuSiQue (Passage Split)
39.46ScoreBaseline
Evaluation results
| Method | Links | |
|---|---|---|
| BaselineModel=ChatGLM, Setting=Passage Split2026.03 | 39.46 | |
| OurModel=ChatGLM, Setting=Passage Split2026.03 | 37.58 | |
| BaselineModel=Qwen, Setting=Passage Split2026.03 | 37.18 | |
| Our + ReorderModel=ChatGLM, Setting=Passage Split2026.03 | 36.35 | |
| OurModel=Qwen, Setting=Passage Split2026.03 | 33.86 | |
| Our + ReorderModel=Qwen, Setting=Passage Split2026.03 | 32.85 | |
| BaselineModel=LLaMA, Setting=Passage Split2026.03 | 32.85 | |
| CacheBlendModel=ChatGLM, Setting=Passage Split2026.03 | 31.88 | |
| EPIC (15%)Model=ChatGLM, Setting=Passage Split2026.03 | 30.91 | |
| OurModel=LLaMA, Setting=Passage Split2026.03 | 29.96 | |
| EPIC (15%)Model=LLaMA, Setting=Passage Split2026.03 | 28.52 | |
| Our + ReorderModel=LLaMA, Setting=Passage Split2026.03 | 27.93 | |
| CacheBlendModel=Qwen, Setting=Passage Split2026.03 | 27.65 | |
| CacheBlendModel=LLaMA, Setting=Passage Split2026.03 | 27.08 | |
| EPIC (15%)Model=Qwen, Setting=Passage Split2026.03 | 24.8 | |
| No RecomputeModel=ChatGLM, Setting=Passage Split2026.03 | 24.74 | |
| No RecomputeModel=LLaMA, Setting=Passage Split2026.03 | 24.62 | |
| No RecomputeModel=Qwen, Setting=Passage Split2026.03 | 10.12 |