Open-domain Question Answering on Natural Questions (test)
48.78EMFull
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FullModel size=7B, Method Category=WR (Vanilla Retrieval)2025.07 | 48.78 | 74.4 | 1 | |
| EviOmniModel size=7B, Method Category=RAR (Reasoning-Augmented)2025.07 | 46.95 | 73.99 | 17.9 | |
| SEERModel size=7B, Method Category=VAR (Context Compression)2025.07 | 45.01 | 72.76 | 12.4 | |
| FilCoModel size=7B, Method Category=VAR (Context Compression)2025.07 | 44.79 | 72.3 | 15.4 | |
| CoTModel size=7B, Method Category=RAR (Reasoning-Augmented)2025.07 | 44.49 | 73.09 | 3.36 | |
| RECOMPModel size=7B, Method Category=VAR (Context Compression)2025.07 | 43.77 | 71.61 | 5.43 | |
| FullModel size=1.5B, Method Category=WR (Vanilla Retrieval)2025.07 | 41.97 | 70.07 | 1 | |
| EviOmniModel size=1.5B, Method Category=RAR (Reasoning-Augmented)2025.07 | 41.14 | 70.77 | 38.1 | |
| CoTModel size=1.5B, Method Category=RAR (Reasoning-Augmented)2025.07 | 37.7 | 68.09 | 4.56 | |
| RECOMPModel size=1.5B, Method Category=VAR (Context Compression)2025.07 | 37.4 | 67.18 | 5.43 | |
| SEERModel size=1.5B, Method Category=VAR (Context Compression)2025.07 | 36.93 | 67.11 | 13.2 | |
| FilCoModel size=1.5B, Method Category=VAR (Context Compression)2025.07 | 36.62 | 66.63 | 16.3 | |
| LLMLingua-2Model size=7B, Method Category=VAR (Context Compression)2025.07 | 36.51 | 67.77 | 4.51 | |
| SeleCtxModel size=7B, Method Category=VAR (Context Compression)2025.07 | 33.91 | 65.71 | 3.44 | |
| LLMLingua-2Model size=1.5B, Method Category=VAR (Context Compression)2025.07 | 29.36 | 62.82 | 4.51 | |
| ZeroModel size=7B, Method Category=WR (Vanilla Retrieval)2025.07 | 25.04 | 60.88 | — | |
| SeleCtxModel size=1.5B, Method Category=VAR (Context Compression)2025.07 | 24.68 | 59.52 | 3.44 | |
| ZeroModel size=1.5B, Method Category=WR (Vanilla Retrieval)2025.07 | 13.74 | 53.16 | — |