Hallucination Detection on MMMU Pro
88.9AUCMHADhs
Evaluation Results
| Method | Links | |
|---|---|---|
| MHADhsVLM Backbone=Llama-3.2-11B2026.04 | 88.9 | |
| MHADffnVLM Backbone=Llama-3.2-11B2026.04 | 87.4 | |
| MHADattnVLM Backbone=Llama-3.2-11B2026.04 | 87 | |
| MINDVLM Backbone=Llama-3.2-11B2026.04 | 86.6 | |
| EnsemHalDetHSVLM Backbone=Llama-3.2-11B2026.04 | 86 | |
| EnsemHalDetAHVLM Backbone=Llama-3.2-11B2026.04 | 85.7 | |
| EnsemHalDetMIXVLM Backbone=Llama-3.2-11B2026.04 | 85.4 | |
| SAPLMAVLM Backbone=Llama-3.2-11B2026.04 | 85 | |
| EnsemHalDetHSVLM Backbone=Pixtral-12B2026.04 | 82.6 | |
| MHADattnVLM Backbone=Pixtral-12B2026.04 | 82.2 | |
| MHADffnVLM Backbone=Pixtral-12B2026.04 | 81 | |
| EnsemHalDetAHVLM Backbone=Pixtral-12B2026.04 | 80.2 | |
| EnsemHalDetMIXVLM Backbone=Pixtral-12B2026.04 | 80.2 | |
| MINDVLM Backbone=Pixtral-12B2026.04 | 80.1 | |
| MHADhsVLM Backbone=Pixtral-12B2026.04 | 79.2 | |
| SAPLMAVLM Backbone=Pixtral-12B2026.04 | 79.1 | |
| EnsemHalDetHSVLM Backbone=Qwen2.5-VL-7B2026.04 | 76.2 | |
| MHADattnVLM Backbone=Qwen2.5-VL-7B2026.04 | 75.1 | |
| EnsemHalDetMIXVLM Backbone=Qwen2.5-VL-7B2026.04 | 74.8 | |
| MHADffnVLM Backbone=Qwen2.5-VL-7B2026.04 | 74.2 | |
| EnsemHalDetAHVLM Backbone=Qwen2.5-VL-7B2026.04 | 74 | |
| MHADhsVLM Backbone=Qwen2.5-VL-7B2026.04 | 70.7 | |
| MINDVLM Backbone=Qwen2.5-VL-7B2026.04 | 68.6 | |
| SAPLMAVLM Backbone=Qwen2.5-VL-7B2026.04 | 61.5 |