Hallucination detection on MSCOCO
91.7AUC ROCHALLUSHIFT++
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HALLUSHIFT++Model=Llama-3.2-11B-Vision2025.12 | 91.7 | 83 | 71.8 | |
| HALLUSHIFT++Model=paligemma-3b-mix-2242025.12 | 87 | 77.5 | 69 | |
| HALLUSHIFT++Model=Llama-3.2-11B-Vision-Instruct2025.12 | 86.1 | 77.4 | 62.6 | |
| HALLUSHIFT++Model=kosmos-2-patch14-2242025.12 | 85.4 | 67.2 | 57.6 | |
| HALLUSHIFT++Model=Phi-3.5-vision-instruct2025.12 | 85.3 | 71.4 | 56.9 | |
| HALLUSHIFT++Model=instructblip-vicuna-7b2025.12 | 85 | 67.7 | 56.4 | |
| HALLUSHIFT++Model=llava-1.5-7b-hf2025.12 | 84.7 | 67.2 | 58.4 | |
| HALLUSHIFT++Model=Qwen2.5-VL-3B-Instruct2025.12 | 83.6 | 72.9 | 58.5 | |
| HALLUSHIFTModel=Llama-3.2-11B-Vision2025.12 | 65.8 | 52.5 | 60.9 | |
| HALLUSHIFTModel=paligemma-3b-mix-2242025.12 | 61.7 | 65.4 | 64.2 | |
| ITIModel=paligemma-3b-mix-2242025.12 | 58.4 | 62.1 | 61.5 | |
| HALLUSHIFTModel=Qwen2.5-VL-3B-Instruct2025.12 | 55.9 | 61.1 | 56.2 | |
| HALLUSHIFTModel=Phi-3.5-vision-instruct2025.12 | 54.9 | 45.9 | 54.3 | |
| HALLUSHIFTModel=kosmos-2-patch14-2242025.12 | 54 | 43.7 | 41 | |
| HALLUSHIFTModel=llava-1.5-7b-hf2025.12 | 53.1 | 42.1 | 43.1 | |
| HALLUSHIFTModel=Llama-3.2-11B-Vision-Instruct2025.12 | 52.1 | 52.1 | 60.5 | |
| HALLUSHIFTModel=instructblip-vicuna-7b2025.12 | 51.7 | 34.3 | 43.3 | |
| ITIModel=kosmos-2-patch14-2242025.12 | 50.3 | 40.8 | 38.2 | |
| ITIModel=Llama-3.2-11B-Vision-Instruct2025.12 | 49.8 | 48.3 | 57.2 |