Hallucination Detection on VQA-RAD (All)
78.23AUCVIHD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VIHDBackbone=Hulu-4B2026.05 | 78.23 | 80.82 | |
| VIHDBackbone=Lingshu-7B2026.05 | 71.58 | 70.86 | |
| VASEBackbone=Lingshu-7B2026.05 | 69.62 | 69.35 | |
| CEBaGMLLM Backbone=HuatuoGPT-Vision-7B2026.03 | 67.7 | 67.6 | |
| CEBaGMLLM Backbone=MedGemma-1.5-4b-it2026.03 | 66.9 | 58.2 | |
| VASEBackbone=Hulu-4B2026.05 | 66.23 | 71.51 | |
| RadFlagBackbone=InternVL3-8B, Uncertainty-based=true2026.01 | 63.99 | 57.98 | |
| AvgProbMLLM Backbone=MedGemma-1.5-4b-it2026.03 | 63.6 | 56.2 | |
| CEBaGMLLM Backbone=LLaVA-Med-v1.5-Mistral-7B2026.03 | 62.1 | 50 | |
| MaxEntBackbone=InternVL3-8B2026.01 | 61.59 | 57.14 | |
| MaxProbBackbone=Hulu-4B2026.05 | 61.17 | 66.98 | |
| MaxEntBackbone=Hulu-4B2026.05 | 61.12 | 66.98 | |
| V-LoopBackbone=InternVL3-8B2026.01 | 60.98 | 58.15 | |
| CEBaGMLLM Backbone=MedGemma-4b-it2026.03 | 60.8 | 64.6 | |
| V-LoopBackbone=MedGemma-4B-it2026.01 | 60.3 | 66.91 | |
| AvgProbBackbone=Hulu-4B2026.05 | 60.13 | 67.02 | |
| SEBackbone=InternVL3-8B, Uncertainty-based=true2026.01 | 60.04 | 56.01 | |
| RadFlagMLLM Backbone=HuatuoGPT-Vision-7B2026.03 | 60 | 63.5 | |
| AvgEntBackbone=InternVL3-8B2026.01 | 59.67 | 54.34 | |
| AvgEntBackbone=Hulu-4B2026.05 | 59.32 | 66.88 | |
| VASEMLLM Backbone=HuatuoGPT-Vision-7B2026.03 | 58.9 | 66.1 | |
| AvgProbMLLM Backbone=LLaVA-Med-v1.5-Mistral-7B2026.03 | 58.8 | 48.7 | |
| RadFlagBackbone=MedGemma-4B-it, Uncertainty-based=true2026.01 | 58.69 | 57.94 | |
| SEMLLM Backbone=HuatuoGPT-Vision-7B2026.03 | 58.3 | 65.1 | |
| SEMLLM Backbone=LLaVA-Med-v1.5-Mistral-7B2026.03 | 58.1 | 50.7 | |
| VASEMLLM Backbone=MedGemma-4b-it2026.03 | 57.6 | 59.9 | |
| VASEMLLM Backbone=LLaVA-Med-v1.5-Mistral-7B2026.03 | 57.6 | 49.3 | |
| RadFlagMLLM Backbone=LLaVA-Med-v1.5-Mistral-7B2026.03 | 56.8 | 48.1 | |
| SEBackbone=MedGemma-4B-it, Uncertainty-based=true2026.01 | 56.65 | 59.76 | |
| MaxEntBackbone=MedGemma-4B-it2026.01 | 56.11 | 58.97 | |
| AvgEntBackbone=MedGemma-4B-it2026.01 | 55.6 | 58.33 | |
| VASEMLLM Backbone=MedGemma-1.5-4b-it2026.03 | 54.8 | 51.5 | |
| AvgEntBackbone=Lingshu-7B2026.05 | 54.8 | 59.34 | |
| MaxEntBackbone=Lingshu-7B2026.05 | 54.74 | 59.33 | |
| AvgProbBackbone=Lingshu-7B2026.05 | 54.72 | 59.29 | |
| MaxProbBackbone=Lingshu-7B2026.05 | 54.71 | 59.3 | |
| RadFlagMLLM Backbone=MedGemma-4b-it2026.03 | 54.7 | 58.9 | |
| RadFlagMLLM Backbone=MedGemma-1.5-4b-it2026.03 | 54.3 | 46.3 | |
| SEMLLM Backbone=MedGemma-1.5-4b-it2026.03 | 53.4 | 48.6 | |
| RadFlagBackbone=Lingshu-7B2026.05 | 52.94 | 58.94 | |
| SEMLLM Backbone=MedGemma-4b-it2026.03 | 52.2 | 58 | |
| MaxEntBackbone=Lingshu-7B2026.01 | 51.76 | 60.96 | |
| AvgEntBackbone=Lingshu-7B2026.01 | 51.73 | 60.96 | |
| RadFlagBackbone=Lingshu-7B, Uncertainty-based=true2026.01 | 51.55 | 59.29 | |
| SEBackbone=Lingshu-7B, Uncertainty-based=true2026.01 | 50.96 | 59.16 | |
| SEBackbone=Lingshu-7B2026.05 | 50.92 | 55.27 | |
| V-LoopBackbone=Lingshu-7B2026.01 | 50.04 | 62.84 | |
| RadFlagBackbone=Hulu-4B2026.05 | 50 | 61.26 | |
| SEBackbone=Hulu-4B2026.05 | 49.48 | 62.24 | |
| AvgProbBackbone=Lingshu-7B2026.01 | 48.27 | 56.36 | |
| MaxProbBackbone=Lingshu-7B2026.01 | 48.24 | 55.44 | |
| AvgProbBackbone=MedGemma-4B-it2026.01 | 44.49 | 54.25 | |
| MaxProbBackbone=MedGemma-4B-it2026.01 | 43.71 | 53.12 | |
| AvgProbMLLM Backbone=MedGemma-4b-it2026.03 | 42.9 | 54.6 | |
| AvgProbBackbone=InternVL3-8B2026.01 | 39.77 | 43.33 | |
| AvgProbMLLM Backbone=HuatuoGPT-Vision-7B2026.03 | 39.3 | 52.2 | |
| MaxProbBackbone=InternVL3-8B2026.01 | 38.66 | 40.72 |