Question Answering on TriviaQA (EM, F1, LLM)
37.74EMSeleCom
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SeleComBackbone=Mistral-7B-Instruct, Top-k=5, Comp. Rate=82x2026.01 | 37.74 | 49.83 | 48.37 | |
| LLM (with RAG*)Backbone=Mistral-7B-Instruct, Top-k=5, Comp. Rate=-2026.01 | 37.71 | 50.19 | 49.06 | |
| LLM (with RAG*)Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=-2026.01 | 35.45 | 48.72 | 46.8 | |
| SeleComBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=82x2026.01 | 35.26 | 47.27 | 45.96 | |
| SeleComBackbone=Qwen2.5-7B-Instruct, Top-k=5, Comp. Rate=82x2026.01 | 33.34 | 45.56 | 44.69 | |
| SeleComBackbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=82x2026.01 | 30.47 | 42.43 | 42.42 | |
| LLM (with RAG*)Backbone=Qwen2.5-7B-Instruct, Top-k=5, Comp. Rate=-2026.01 | 28.95 | 44.76 | 46.32 | |
| LLM (with RAG*)Backbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=-2026.01 | 26.49 | 41.12 | 41.07 | |
| COCOMBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=128x2026.01 | 13.75 | 30.68 | 37.58 | |
| LLM (with RAG)Backbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=-2026.01 | 9.63 | 24.13 | 37.68 | |
| ICAEBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=4x2026.01 | 9.53 | 25.03 | 38.12 | |
| RECOMPBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=N/A2026.01 | 7.43 | 22.13 | 41.45 | |
| LLM (w/o RAG)Backbone=Qwen2.5-7B-Instruct, Top-k=1, Comp. Rate=-2026.01 | 6.96 | 19.46 | 28.6 | |
| LLM (with RAG)Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=-2026.01 | 6.41 | 19.81 | 40 | |
| LLMLingua-2Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=4x2026.01 | 5.72 | 18.34 | 37.21 | |
| LLM (w/o RAG)Backbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=-2026.01 | 5.57 | 19.18 | 38.93 | |
| xRAGBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=164x2026.01 | 5.37 | 17.92 | 40.28 | |
| PISCOBackbone=Mistral-7B-Instruct, Top-k=1, Comp. Rate=64x2026.01 | 0.8 | 9.11 | 42.11 |