Hallucination Detection on CRAG-MM
0.874AUCEnsemHalDetMIX
Evaluation Results
| Method | Links | |
|---|---|---|
| EnsemHalDetMIXVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.874 | |
| EnsemHalDetAHVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.871 | |
| MHADattnVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.867 | |
| EnsemHalDetHSVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.864 | |
| EnsemHalDetHSVLM Backbone=Llama-3.2-11B2026.04 | 0.86 | |
| MHADffnVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.855 | |
| MHADattnVLM Backbone=Llama-3.2-11B2026.04 | 0.85 | |
| MHADhsVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.849 | |
| EnsemHalDetMIXVLM Backbone=Llama-3.2-11B2026.04 | 0.849 | |
| EnsemHalDetAHVLM Backbone=Llama-3.2-11B2026.04 | 0.843 | |
| MHADhsVLM Backbone=Llama-3.2-11B2026.04 | 0.838 | |
| MINDVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.83 | |
| MHADffnVLM Backbone=Llama-3.2-11B2026.04 | 0.83 | |
| MINDVLM Backbone=Llama-3.2-11B2026.04 | 0.822 | |
| EnsemHalDetMIXVLM Backbone=Pixtral-12B2026.04 | 0.818 | |
| SAPLMAVLM Backbone=Llama-3.2-11B2026.04 | 0.817 | |
| EnsemHalDetHSVLM Backbone=Pixtral-12B2026.04 | 0.817 | |
| EnsemHalDetAHVLM Backbone=Pixtral-12B2026.04 | 0.8 | |
| SAPLMAVLM Backbone=Qwen2.5-VL-7B2026.04 | 0.797 | |
| MHADattnVLM Backbone=Pixtral-12B2026.04 | 0.784 | |
| MHADffnVLM Backbone=Pixtral-12B2026.04 | 0.777 | |
| MHADhsVLM Backbone=Pixtral-12B2026.04 | 0.763 | |
| MINDVLM Backbone=Pixtral-12B2026.04 | 0.75 | |
| SAPLMAVLM Backbone=Pixtral-12B2026.04 | 0.73 |