Hallucination Detection on VQA-Med Open-Ended 2019
78.65AUCVIHD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VIHDBackbone=Hulu-4B2026.05 | 78.65 | 59.64 | |
| VASEBackbone=Hulu-4B2026.05 | 75.09 | 54.3 | |
| VIHDBackbone=Lingshu-7B2026.05 | 74.71 | 64.11 | |
| V-LoopBackbone=InternVL3-8B2026.01 | 72.86 | 56.58 | |
| V-LoopBackbone=Lingshu-7B2026.01 | 71.93 | 66.33 | |
| MaxEntBackbone=Hulu-4B2026.05 | 71.67 | 52.74 | |
| MaxProbBackbone=Hulu-4B2026.05 | 71.45 | 51.86 | |
| AvgProbBackbone=Hulu-4B2026.05 | 70.48 | 51.56 | |
| AvgEntBackbone=Hulu-4B2026.05 | 70.43 | 51.68 | |
| VASEBackbone=Lingshu-7B2026.05 | 69.57 | 56.77 | |
| RadFlagBackbone=InternVL3-8B, Uncertainty-based=true2026.01 | 67.69 | 53.58 | |
| SEBackbone=InternVL3-8B, Uncertainty-based=true2026.01 | 66.18 | 52.19 | |
| V-LoopBackbone=MedGemma-4B-it2026.01 | 66.02 | 50.92 | |
| RadFlagBackbone=MedGemma-4B-it, Uncertainty-based=true2026.01 | 64.91 | 36.42 | |
| MaxEntBackbone=InternVL3-8B2026.01 | 63.81 | 46.94 | |
| SEBackbone=MedGemma-4B-it, Uncertainty-based=true2026.01 | 61.88 | 47.02 | |
| AvgEntBackbone=InternVL3-8B2026.01 | 60.46 | 42.05 | |
| MaxEntBackbone=MedGemma-4B-it2026.01 | 58.58 | 33.58 | |
| RadFlagBackbone=Lingshu-7B2026.05 | 57.3 | 35.91 | |
| AvgEntBackbone=MedGemma-4B-it2026.01 | 56.94 | 32.85 | |
| AvgEntBackbone=Lingshu-7B2026.05 | 55.68 | 35.63 | |
| MaxEntBackbone=Lingshu-7B2026.05 | 55.64 | 35.59 | |
| AvgProbBackbone=Lingshu-7B2026.05 | 55.61 | 35.54 | |
| MaxProbBackbone=Lingshu-7B2026.05 | 55.59 | 35.54 | |
| RadFlagBackbone=Lingshu-7B, Uncertainty-based=true2026.01 | 55.24 | 40.67 | |
| SEBackbone=Lingshu-7B2026.05 | 54.27 | 34.69 | |
| MaxEntBackbone=Lingshu-7B2026.01 | 54.13 | 40 | |
| AvgEntBackbone=Lingshu-7B2026.01 | 54.07 | 39.98 | |
| SEBackbone=Lingshu-7B, Uncertainty-based=true2026.01 | 51.8 | 39.39 | |
| RadFlagBackbone=Hulu-4B2026.05 | 50 | 21.74 | |
| SEBackbone=Hulu-4B2026.05 | 47.53 | 19.58 | |
| AvgProbBackbone=Lingshu-7B2026.01 | 45.93 | 46.15 | |
| MaxProbBackbone=Lingshu-7B2026.01 | 45.87 | 44.81 | |
| AvgProbBackbone=MedGemma-4B-it2026.01 | 42.96 | 35.43 | |
| MaxProbBackbone=MedGemma-4B-it2026.01 | 41.28 | 30 | |
| AvgProbBackbone=InternVL3-8B2026.01 | 38.89 | 33.64 | |
| MaxProbBackbone=InternVL3-8B2026.01 | 36.44 | 28.27 |