Medical Search Relevance Assessment on LocalQSMed HARD v1.0
0.6611AccuracyQwen3-32B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-32BLoss Rate=0.00%2025.12 | 0.6611 | |
| Qwen3-14BLoss Rate=0.02%2025.12 | 0.6106 | |
| DeepSeek-R1-Distill-Qwen-14BLoss Rate=0.00%2025.12 | 0.5758 | |
| Qwen3-8BLoss Rate=0.00%2025.12 | 0.549 | |
| Qwen3-4BLoss Rate=0.00%2025.12 | 0.5294 | |
| Llama-3.3-70B-InstructLoss Rate=46.86%2025.12 | 0.3737 | |
| DeepSeek-R1-Distill-Llama-70BLoss Rate=43.68%2025.12 | 0.3598 | |
| Llama-3.1-8BLoss Rate=63.55%2025.12 | 0.3504 | |
| Llama-3.2-3BLoss Rate=0.00%2025.12 | 0.3358 | |
| Llama-3.2-1BLoss Rate=63.67%2025.12 | 0.2913 | |
| Qwen3-0.6BLoss Rate=6.40%2025.12 | 0.2784 | |
| DeepSeek-R1-Distill-Qwen-7BLoss Rate=17.59%2025.12 | 0.0699 |