Object Hallucination Detection on POPE MS-COCO Overall
86.83AccuracyBRACS
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BRACSBackbone=Qwen-VL-Chat, Decoding Strategy=BRACS2026.05 | 86.83 | 85.81 | |
| VDD-NoneBackbone=Qwen-VL-Chat, Decoding Strategy=VDD-None2026.05 | 86.78 | 85.78 | |
| BRACSBackbone=LLaVA-1.5-7B, Decoding Strategy=BRACS2026.05 | 86.63 | 86.37 | |
| PAIBackbone=Qwen-VL-Chat, Decoding Strategy=PAI2026.05 | 86.26 | 85.33 | |
| SPINBackbone=LLaVA-1.5-7B, Decoding Strategy=SPIN2026.05 | 85.5 | 84.15 | |
| VCDBackbone=Qwen-VL-Chat, Decoding Strategy=VCD2026.05 | 85.4 | 83.95 | |
| SPINBackbone=Qwen-VL-Chat, Decoding Strategy=SPIN2026.05 | 85.19 | 83.53 | |
| GreedyBackbone=LLaVA-1.5-7B, Decoding Strategy=Greedy2026.05 | 85.13 | 83.64 | |
| GreedyBackbone=Qwen-VL-Chat, Decoding Strategy=Greedy2026.05 | 85.1 | 83.42 | |
| VDD-NoneBackbone=LLaVA-1.5-7B, Decoding Strategy=VDD-None2026.05 | 84.51 | 85.55 | |
| VCDBackbone=LLaVA-1.5-7B, Decoding Strategy=VCD2026.05 | 84.21 | 83.19 | |
| PAIBackbone=LLaVA-1.5-7B, Decoding Strategy=PAI2026.05 | 84.02 | 85.3 |