Medical Question Answering on MedicalMeadow WikiDoc PatientInfo Med
20.5MPADMI
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ADMIBackbone=Llama3-8B2026.06 | 20.5 | 0.7 | 99.5 | |
| DPForwardBackbone=Llama3-8B2026.06 | 19.4 | 84.5 | 62.6 | |
| MIDBackbone=Qwen2.5-7B2026.06 | 19.1 | 21.1 | 87 | |
| w/o defenseBackbone=Llama3-8B2026.06 | 18.7 | 88.1 | — | |
| CuxTextBackbone=Llama3.2-3B2026.06 | 18.5 | 84.1 | 62.7 | |
| w/o defenseBackbone=Qwen2.5-7B2026.06 | 18.4 | 80.1 | — | |
| ADMIBackbone=Qwen2.5-7B2026.06 | 18.2 | 0.7 | 99.5 | |
| DualGuardBackbone=Qwen2.5-7B2026.06 | 17.8 | 50.3 | 73.8 | |
| w/o defenseBackbone=Llama3.2-3B2026.06 | 17.6 | 90.1 | — | |
| ADMIBackbone=Llama3.2-3B2026.06 | 17.3 | 7.4 | 95 | |
| MIDBackbone=Llama3.2-3B2026.06 | 16.9 | 13.8 | 91.1 | |
| SPBackbone=Qwen2.5-7B2026.06 | 16.8 | 68.6 | 67.3 | |
| CuxTextBackbone=Llama3-8B2026.06 | 16.6 | 90 | 61.1 | |
| MIDBackbone=Llama3-8B2026.06 | 16.5 | 2.7 | 97.6 | |
| SPBackbone=Llama3.2-3B2026.06 | 16.2 | 53.8 | 72.4 | |
| SPBackbone=Llama3-8B2026.06 | 16.2 | 67.2 | 67.8 | |
| DPForwardBackbone=Llama3.2-3B2026.06 | 16.1 | 68 | 67.5 | |
| SanTextBackbone=Qwen2.5-7B2026.06 | 14.9 | 73 | 65.9 | |
| DualGuardBackbone=Llama3.2-3B2026.06 | 14.7 | 31.2 | 81.9 | |
| DualGuardBackbone=Llama3-8B2026.06 | 14.4 | 44.9 | 75.8 | |
| SanTextBackbone=Llama3.2-3B2026.06 | 14.1 | 79.6 | 64 | |
| DPForwardBackbone=Qwen2.5-7B2026.06 | 13.8 | 72.9 | 65.9 | |
| SanTextBackbone=Llama3-8B2026.06 | 10.5 | 90 | 61 | |
| CuxTextBackbone=Qwen2.5-7B2026.06 | 9.7 | 62.7 | 69.1 |