Hallucination Detection on HalLoc-VQA
77.9ROCFaithSCAN
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FaithSCANBackbone=LLaVA-8B2026.01 | 77.9 | 82.3 | 68 | 87.3 | 1,898.95 | 0.741 | |
| SEUBackbone=LLaVA-8B2026.01 | 76.9 | 79.9 | 66.7 | 86.5 | 2,443.89 | 1.769 | |
| FaithSCANBackbone=IB-T5-XL2026.01 | 75.4 | 77.8 | 75.5 | 79.2 | 1,560.237 | 0.605 | |
| P(True)Backbone=Qwen3-VL-8B2026.01 | 75 | 82.5 | 54.8 | 83.8 | 1,143.72 | 24.299 | |
| EmbVarBackbone=LLaVA-8B2026.01 | 74.7 | 82 | 50.9 | 85.5 | 2,504.584 | 0.175 | |
| FaithSCANBackbone=Qwen3-VL-8B2026.01 | 73.9 | 75.7 | 56.9 | 81.1 | 1,741.652 | 1.191 | |
| LogisticBackbone=LLaVA-8B2026.01 | 73.3 | 78.8 | 58 | 83 | 1,874.895 | 0.114 | |
| SelfCheckGPTBackbone=LLaVA-8B2026.01 | 71.5 | 77.9 | 62.5 | 81.8 | 2,588.092 | 340.683 | |
| PEBackbone=Qwen3-VL-8B2026.01 | 68.6 | 17.8 | 53.5 | 20 | 1,328.434 | 0.002 | |
| SEBackbone=LLaVA-8B2026.01 | 68.3 | 74.7 | 46.2 | 79.5 | 2,632.833 | 11.212 | |
| P(True)Backbone=IB-T5-XL2026.01 | 67.5 | 59.7 | 75.3 | 56.5 | 1,127.961 | 14.938 | |
| LogisticBackbone=Qwen3-VL-8B2026.01 | 67.2 | 68.6 | 49.6 | 72.4 | 1,689.458 | 0.102 | |
| LogisticBackbone=IB-T5-XL2026.01 | 66.9 | 66.9 | 73.4 | 67.4 | 1,570.349 | 0.087 | |
| SEUBackbone=IB-T5-XL2026.01 | 66.3 | 45.2 | 73.8 | 44 | 1,685.975 | 1.845 | |
| EmbVarBackbone=IB-T5-XL2026.01 | 65.7 | 46.4 | 74.6 | 45.3 | 1,430.884 | 0.06 | |
| SelfCheckGPTBackbone=Qwen3-VL-8B2026.01 | 65.6 | 71.6 | 55.2 | 74.8 | 2,165.013 | 367.522 | |
| SEUBackbone=Qwen3-VL-8B2026.01 | 65.3 | 74.2 | 58.7 | 80 | 2,199.427 | 1.821 | |
| T-NLLBackbone=Qwen3-VL-8B2026.01 | 65 | 23.3 | 52.4 | 19 | 1,330.996 | 0.001 | |
| EmbVarBackbone=Qwen3-VL-8B2026.01 | 62.8 | 70.5 | 57 | 78.7 | 2,194.586 | 0.107 | |
| P(True)Backbone=LLaVA-8B2026.01 | 61.5 | 69 | 52 | 74.8 | 1,178.084 | 19.511 | |
| T-NLLBackbone=LLaVA-8B2026.01 | 60.7 | 25.6 | 52.9 | 24.7 | 1,524.74 | 0.001 | |
| PEBackbone=LLaVA-8B2026.01 | 60.5 | 27.6 | 53.5 | 25.8 | 1,491.448 | 0.002 | |
| SEBackbone=IB-T5-XL2026.01 | 58.3 | 56.3 | 65.3 | 56.2 | 1,381.929 | 4.913 | |
| SEBackbone=Qwen3-VL-8B2026.01 | 55.9 | 32.5 | 45.6 | 30.5 | 2,019.877 | 13.812 | |
| PEBackbone=IB-T5-XL2026.01 | 53.6 | 52.6 | 73 | 52.7 | 1,326.308 | 0.002 | |
| T-NLLBackbone=IB-T5-XL2026.01 | 53.5 | 53.7 | 76.2 | 54.2 | 1,257.512 | 0.001 | |
| SelfCheckGPTBackbone=IB-T5-XL2026.01 | 50.9 | 54.5 | 72.2 | 59 | 1,602.887 | 78.613 |