Question Answering on RepliQA 1,000 QA pairs (test)
37AccuracyActIdx
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ActIdxModel=Q-14B2025.06 | 37 | 30.8 | 31.5 | |
| ActIdxModel=L-8B2025.06 | 35.7 | 36.2 | 36.9 | |
| ActIdxModel=Q-7B2025.06 | 31.9 | 24.4 | 25.7 | |
| PassIdxModel=Q-14B2025.06 | 28.4 | 2 | 2 | |
| InsOnlyModel=Q-14B2025.06 | 27.6 | 0.75 | 0.9 | |
| ActIdxModel=L-3B2025.06 | 26.8 | 13.9 | 23 | |
| PassIdxModel=L-8B2025.06 | 25.6 | 1.9 | 1.9 | |
| PassIdxModel=Q-7B2025.06 | 24.8 | 2.4 | 2.5 | |
| ActIdxModel=Q-3B2025.06 | 24.5 | 10.5 | 15.8 | |
| InsOnlyModel=Q-7B2025.06 | 24.2 | 0.9 | 1.3 | |
| InsOnlyModel=L-8B2025.06 | 24.1 | 0.7 | 1.4 | |
| PassIdxModel=L-3B2025.06 | 23.4 | 1.8 | 2 | |
| PassIdxModel=Q-3B2025.06 | 22.8 | 1.8 | 2.4 | |
| InsOnlyModel=L-3B2025.06 | 19 | 0.27 | 0.4 | |
| InsOnlyModel=Q-3B2025.06 | 15.2 | 0.2 | 0.2 |