Long-context Question Answering on NarrativeQA Passage Split
32.64ScoreBaseline
Evaluation results
| Method | Links | |
|---|---|---|
| BaselineModel=ChatGLM, Setting=Passage Split2026.03 | 32.64 | |
| Our + ReorderModel=LLaMA, Setting=Passage Split2026.03 | 32.43 | |
| No RecomputeModel=ChatGLM, Setting=Passage Split2026.03 | 31.81 | |
| Our + ReorderModel=ChatGLM, Setting=Passage Split2026.03 | 31.8 | |
| CacheBlendModel=ChatGLM, Setting=Passage Split2026.03 | 31.79 | |
| EPIC (15%)Model=ChatGLM, Setting=Passage Split2026.03 | 31.42 | |
| OurModel=LLaMA, Setting=Passage Split2026.03 | 31.41 | |
| EPIC (15%)Model=LLaMA, Setting=Passage Split2026.03 | 31.02 | |
| OurModel=ChatGLM, Setting=Passage Split2026.03 | 31 | |
| CacheBlendModel=LLaMA, Setting=Passage Split2026.03 | 30.95 | |
| No RecomputeModel=LLaMA, Setting=Passage Split2026.03 | 28.1 | |
| Our + ReorderModel=Qwen, Setting=Passage Split2026.03 | 23.1 | |
| EPIC (15%)Model=Qwen, Setting=Passage Split2026.03 | 22.91 | |
| OurModel=Qwen, Setting=Passage Split2026.03 | 22.88 | |
| CacheBlendModel=Qwen, Setting=Passage Split2026.03 | 21.97 | |
| No RecomputeModel=Qwen, Setting=Passage Split2026.03 | 20.78 | |
| BaselineModel=LLaMA, Setting=Passage Split2026.03 | 18.62 | |
| BaselineModel=Qwen, Setting=Passage Split2026.03 | 16.54 |