Object Hallucination Detection on POPE (Popular Average across MS-COCO, A-OKVQA, GQA)
85.2AccuracyVAF
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VAFBackbone=LLaVA-v1.5-13B2025.03 | 85.2 | 86.4 | |
| VAFBackbone=Qwen-VL-Chat-7B2025.03 | 84.9 | 85.1 | |
| VAFBackbone=LLaVA-v1.5-7B2025.03 | 84.5 | 84.9 | |
| VCDBackbone=LLaVA-v1.5-13B2025.03 | 83.7 | 85.1 | |
| ICDBackbone=Qwen-VL-Chat-7B2025.03 | 83.2 | 84.5 | |
| VCDBackbone=LLaVA-v1.5-7B2025.03 | 83.1 | 84.1 | |
| VCDBackbone=Qwen-VL-Chat-7B2025.03 | 83 | 84.1 | |
| ICDBackbone=LLaVA-v1.5-13B2025.03 | 82.9 | 84.3 | |
| RegularBackbone=LLaVA-v1.5-13B2025.03 | 82.7 | 84.1 | |
| RegularBackbone=LLaVA-v1.5-7B2025.03 | 82.5 | 83.2 | |
| RegularBackbone=Qwen-VL-Chat-7B2025.03 | 82.4 | 83.1 | |
| ICDBackbone=LLaVA-v1.5-7B2025.03 | 82.1 | 82.9 |