Long-context Question Answering on HotpotQA Passage
60.03ScoreBaseline
Evaluation results
| Method | Links | |
|---|---|---|
| BaselineModel=ChatGLM, Setting=Passage Split2026.03 | 60.03 | |
| Our + ReorderModel=Qwen, Setting=Passage Split2026.03 | 59.72 | |
| OurModel=Qwen, Setting=Passage Split2026.03 | 59.54 | |
| BaselineModel=Qwen, Setting=Passage Split2026.03 | 59.22 | |
| CacheBlendModel=Qwen, Setting=Passage Split2026.03 | 57.38 | |
| OurModel=ChatGLM, Setting=Passage Split2026.03 | 56.14 | |
| Our + ReorderModel=ChatGLM, Setting=Passage Split2026.03 | 55.67 | |
| EPIC (15%)Model=ChatGLM, Setting=Passage Split2026.03 | 54.81 | |
| EPIC (15%)Model=Qwen, Setting=Passage Split2026.03 | 54.43 | |
| BaselineModel=LLaMA, Setting=Passage Split2026.03 | 54.1 | |
| Our + ReorderModel=LLaMA, Setting=Passage Split2026.03 | 52.02 | |
| CacheBlendModel=ChatGLM, Setting=Passage Split2026.03 | 51.64 | |
| OurModel=LLaMA, Setting=Passage Split2026.03 | 51.23 | |
| EPIC (15%)Model=LLaMA, Setting=Passage Split2026.03 | 48.98 | |
| CacheBlendModel=LLaMA, Setting=Passage Split2026.03 | 48.72 | |
| No RecomputeModel=ChatGLM, Setting=Passage Split2026.03 | 43.88 | |
| No RecomputeModel=LLaMA, Setting=Passage Split2026.03 | 42.53 | |
| No RecomputeModel=Qwen, Setting=Passage Split2026.03 | 30.59 |