Laboratory-language grounding on LabGuard-Bench (random 80/20 split)
91.5Type Accuracy
Qwen3-8B
Updated 25d ago
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
2026.06 | 91.5 | 97.3 | 93.6 | 86.8 | 72.9 | 100 | 100 | |
2026.06 | 91.4 | 97.1 | 93.5 | 86.6 | 73 | 100 | 100 | |
2026.06 | 77.4 | 89 | 69.8 | 29.1 | 12 | 98.2 | 98.2 | |
2026.06 | 76.8 | 88.5 | 68.9 | 27.3 | 11.5 | 97.6 | 97.6 | |
2026.06 | 74.4 | 87.1 | 66.8 | 24.7 | 9.8 | 96.3 | 96.3 | |
2026.06 | 70.7 | 95.9 | 48.1 | 19.9 | 0.8 | 100 | 100 | |
2026.06 | 68.3 | 85.4 | 63.6 | 20.3 | 7.3 | 95.1 | 95.1 | |
2026.06 | 64 | 82.1 | 55.8 | 12.6 | 5.1 | 88.4 | 88.4 | |
2026.06 | 19.5 | 95.9 | 17.3 | 0.5 | 0.3 | 46.3 | 46.3 |