Hallucination Detection on MedHallu (test)
88.2Precision (HP)GPT-4o
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GPT-4o2026.05 | 88.2 | — | 87.7 | — | — | — | — | |
| CALAMRFLOWAveraged over=5 random seeds2026.05 | 87.4 | 91.1 | 89.2 | 89 | 89 | 89 | — | |
| Qwen2.5-7BParameters=7B2026.05 | 86.6 | — | 83.9 | — | — | — | — | |
| DeepSeek-R1-8BParameters=8B2026.05 | 86.4 | — | 81.2 | — | — | — | — | |
| Qwen2.5-14BParameters=14B2026.05 | 85.7 | — | 85.2 | — | — | — | — | |
| GPT-4o mini2026.05 | 82 | — | 84.1 | — | — | — | — | |
| Gemma-2-9BParameters=9B2026.05 | 80.9 | — | 83.8 | — | — | — | — | |
| Gemma-2-2BParameters=2B2026.05 | 78.6 | — | 71.5 | — | — | — | — | |
| Llama-3.1-8BParameters=8B2026.05 | 77.5 | — | 79.7 | — | — | — | — | |
| Llama-3.2-3BParameters=3B2026.05 | 77.5 | — | 73.4 | — | — | — | — | |
| Qwen2.5-3BParameters=3B2026.05 | 51.4 | — | 67.6 | — | — | — | — | |
| CALAMRFLOWseeds=5, test-set size (N)=14,0402026.05 | — | — | — | 89 | 89 | — | 89.2 | |
| DeepSeek-R1-Distill-8Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 81.2 | |
| Gemma-2-2Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 71.5 | |
| Gemma-2-9Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 83.8 | |
| GPT-4otest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 87.7 | |
| GPT-4o minitest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 84.1 | |
| Llama-3.1-8Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 79.7 | |
| Llama-3.2-3Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 73.4 | |
| Qwen2.5-14Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 85.2 | |
| Qwen2.5-3Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 67.6 | |
| Qwen2.5-7Btest-set size (N)=14,0402026.05 | — | — | — | — | — | — | 83.9 |