Question Answering on AVG.
47.5EMEXIT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EXITBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 47.5 | 56.9 | |
| Original DocsBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=-2024.12 | 47 | 56.9 | |
| EXITBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 45.7 | 55.1 | |
| LongLLMLinguaBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 45.2 | 54.6 | |
| DecoupleSearchBackbone=Qwen-2.5-14B-Instruct2025.09 | 44.63 | 42.5 | |
| Original DocsBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=-2024.12 | 43.8 | 53.1 | |
| CompActBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 43.7 | 53 | |
| LongLLMLinguaBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 43.7 | 52.8 | |
| RefinerBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 43.1 | 52.4 | |
| RefinerBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 41.6 | 50.3 | |
| RECOMP-ExtrBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 41.4 | 49.7 | |
| RECOMP-AbstBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 41.4 | 50.8 | |
| CompActBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 40.8 | 49.2 | |
| GreedyAgentBackbone=Qwen-2.5-14B-Instruct2025.09 | 40.36 | 39.63 | |
| EXITBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 40.3 | 49.8 | |
| RECOMP-ExtrBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 40.1 | 48.2 | |
| RECOMP-AbstBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 40 | 49.3 | |
| DecoupleSearchBackbone=Qwen-2.5-7B-Instruct2025.09 | 39.93 | 38.79 | |
| Original DocsBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=-2024.12 | 39.7 | 49.4 | |
| LongLLMLinguaBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 38.5 | 47.5 | |
| RAgentBackbone=Qwen-2.5-14B-Instruct2025.09 | 38.23 | 38.86 | |
| EXITBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 38.1 | 47.4 | |
| Original DocsBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=-2024.12 | 37.5 | 47 | |
| Search-o1Backbone=Qwen-2.5-14B-Instruct2025.09 | 37.46 | 39.49 | |
| CompActBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 37.3 | 46.8 | |
| RefinerBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 37.1 | 46.5 | |
| CompActBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 36.9 | 46.3 | |
| RECOMP-AbstBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 36.2 | 45.3 | |
| RECOMP-AbstBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 35.9 | 44.6 | |
| GreedyAgentBackbone=Qwen-2.5-7B-Instruct2025.09 | 35.87 | 35.34 | |
| RefinerBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 35.6 | 44.7 | |
| LongLLMLinguaBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 35.5 | 44.1 | |
| RECOMP-ExtrBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 35.4 | 43.7 | |
| RECOMP-ExtrBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 34 | 42.1 | |
| Search-o1Backbone=Qwen-2.5-7B-Instruct2025.09 | 31.37 | 33.4 | |
| CoTBackbone=Qwen-2.5-14B-Instruct2025.09 | 30.89 | 34.04 | |
| Gen-RetrieveBackbone=Qwen-2.5-14B-Instruct2025.09 | 29.87 | 27.75 | |
| RAgentBackbone=Qwen-2.5-7B-Instruct2025.09 | 29.85 | 32.02 | |
| Jud-RetrieveBackbone=Qwen-2.5-14B-Instruct2025.09 | 29.75 | 27.29 | |
| StandardBackbone=Qwen-2.5-14B-Instruct2025.09 | 29.63 | 27.36 | |
| StandardBackbone=Qwen-2.5-7B-Instruct2025.09 | 25.27 | 27.42 | |
| Gen-RetrieveBackbone=Qwen-2.5-7B-Instruct2025.09 | 25.19 | 27.02 | |
| Judge-RetrieveBackbone=Qwen-2.5-7B-Instruct2025.09 | 25.07 | 27.54 | |
| DirectBackbone=Qwen-2.5-14B-Instruct2025.09 | 24.14 | 26.91 | |
| CoTBackbone=Qwen-2.5-7B-Instruct2025.09 | 22.81 | 26.32 | |
| DirectBackbone=Qwen-2.5-7B-Instruct2025.09 | 21.99 | 24.7 | |
| IterativeBackbone=Qwen-2.5-7B-Instruct2025.09 | 17.15 | 18.93 | |
| IterativeBackbone=Qwen-2.5-14B-Instruct2025.09 | 15.66 | 16.63 |