Question Answering on SciQAG 1,000 sampled questions (test)
70.2AccuracyActIdx
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ActIdxModel=Q-14B2025.06 | 70.2 | 34.8 | 35.1 | |
| ActIdxModel=Q-3B2025.06 | 67.9 | 20 | 23 | |
| ActIdxModel=L-8B2025.06 | 67.7 | 39.3 | 40.3 | |
| InsOnlyModel=Q-14B2025.06 | 67.4 | 0.5 | 0.8 | |
| PassIdxModel=Q-14B2025.06 | 67.4 | 2.3 | 2.3 | |
| InsOnlyModel=L-8B2025.06 | 67.3 | 1 | 1.6 | |
| PassIdxModel=Q-3B2025.06 | 67 | 1.1 | 1.6 | |
| ActIdxModel=Q-7B2025.06 | 66.6 | 32.6 | 33.6 | |
| InsOnlyModel=Q-7B2025.06 | 65.9 | 0.6 | 0.8 | |
| PassIdxModel=Q-7B2025.06 | 65.7 | 2.4 | 2.4 | |
| ActIdxModel=L-3B2025.06 | 65.6 | 13.2 | 32.8 | |
| PassIdxModel=L-8B2025.06 | 65.1 | 3.6 | 3.7 | |
| PassIdxModel=L-3B2025.06 | 63.2 | 1.8 | 2.1 | |
| InsOnlyModel=L-3B2025.06 | 61.8 | 0.2 | 0.2 | |
| InsOnlyModel=Q-3B2025.06 | 61.2 | 0 | 0 |