Hallucination Detection on MRI
96.81HDRCounterVHD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CounterVHDParam.=2B, Backbone=Gemini-32026.06 | 96.81 | 90.1 | |
| CounterVHDParam.=2B, Backbone=Grok-42026.06 | 95.83 | 93.88 | |
| CounterVHDParam.=2B, Backbone=MedGemma2026.06 | 95.79 | 91.92 | |
| CounterVHDParam.=2B, Backbone=GPT-5.12026.06 | 91.93 | 92 | |
| UniHDParam.=-, Backbone=MedGemma2026.06 | 91.58 | 48.33 | |
| UniHDParam.=-, Backbone=Grok-42026.06 | 86.46 | 47.43 | |
| GPT-5.1-basedParam.=-, Backbone=GPT-5.12026.06 | 80.2 | 75 | |
| UniHDParam.=-, Backbone=Gemini-32026.06 | 79.79 | 45.18 | |
| GPT-5.1-basedParam.=-, Backbone=Gemini-32026.06 | 77.49 | 69.77 | |
| GPT-5.1-basedParam.=-, Backbone=Grok-42026.06 | 68.75 | 62 | |
| UniHDParam.=-, Backbone=GPT-5.12026.06 | 65.66 | 45.46 | |
| GPT-5.1-basedParam.=-, Backbone=MedGemma2026.06 | 65.64 | 59.33 | |
| FaithscoreParam.=13B, Backbone=MedGemma2026.06 | 46.32 | 67.69 | |
| FaithscoreParam.=13B, Backbone=GPT-5.12026.06 | 44.44 | 65.67 | |
| FaithscoreParam.=13B, Backbone=Grok-42026.06 | 41.67 | 63.49 | |
| FaithscoreParam.=13B, Backbone=Gemini-32026.06 | 34.04 | 66.67 |