Hallucination Evaluation on MOH
69.5HR^Dbaseline
Evaluation results
| Method | Links | ||
|---|---|---|---|
| baselineModel=LLaVA-v1.5-13B2026.02 | 69.5 | 67.9 | |
| SENTINELModel=LLaVA-v1.5-13B2026.02 | 56.8 | 36.5 | |
| baselineBase Model=LLaVA-v1.5-7B2026.02 | 52.2 | 68.6 | |
| POVIDBase Model=LLaVA-v1.5-7B2026.02 | 51.1 | 65.1 | |
| mDPOBase Model=LLaVA-v1.5-7B2026.02 | 50.7 | 53.2 | |
| Re-AlignBase Model=LLaVA-v1.5-7B2026.02 | 49.9 | 49 | |
| vanilla-DPOModel=LLaVA-v1.5-13B2026.02 | 46.2 | 35 | |
| mDPOModel=LLaVA-v1.5-13B2026.02 | 45.6 | 48.3 | |
| HSA-DPOModel=LLaVA-v1.5-13B2026.02 | 45.3 | 28.2 | |
| HA-DPOBase Model=LLaVA-v1.5-7B2026.02 | 45.2 | 63.4 | |
| HII-DPOModel=LLaVA-v1.5-13B2026.02 | 43.8 | 17.5 | |
| Re-AlignModel=LLaVA-v1.5-13B2026.02 | 39.4 | 51.1 | |
| SENTINELBase Model=LLaVA-v1.5-7B2026.02 | 38.3 | 38.8 | |
| TPOBase Model=LLaVA-v1.5-7B2026.02 | 31.7 | 36.6 | |
| HII-DPOBase Model=LLaVA-v1.5-7B2026.02 | 28.1 | 18.8 | |
| Qwen2-VL-7B-InstructMethod=baseline2026.02 | 11.5 | 58 | |
| HII-DPOModel=Qwen2-VL-7B-Instruct2026.02 | 9 | 20 | |
| HII-DPOModel=Qwen2-VL-2B-Instruct2026.02 | 4.7 | 15.5 | |
| Qwen2-VL-2B-InstructMethod=baseline2026.02 | 4.1 | 59 | |
| Qwen2.5-VL-7B-InstructMethod=baseline2026.02 | 3.6 | 47 | |
| HII-DPOModel=Qwen2.5-VL-7B-Instruct2026.02 | 2.2 | 17.4 |