Polling-based Object Probing Evaluation (POPE) on A-OKVQA (Popular)
0.8813AccuracyICT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ICTModel=Qwen-VL2024.11 | 0.8813 | 0.8783 | |
| OPERAModel=Qwen-VL2024.11 | 0.8791 | 0.8713 | |
| ICT w/o objectModel=Qwen-VL2024.11 | 0.879 | 0.8748 | |
| VCDModel=Qwen-VL2024.11 | 0.8785 | 0.8781 | |
| ICT w/o imageModel=Qwen-VL2024.11 | 0.877 | 0.8768 | |
| ICTModel=LLaVA-v1.52024.11 | 0.8573 | 0.8534 | |
| RegularModel=Qwen-VL2024.11 | 0.8556 | 0.8463 | |
| ICT w/o imageModel=LLaVA-v1.52024.11 | 0.8536 | 0.8534 | |
| OPERAModel=LLaVA-v1.52024.11 | 0.8322 | 0.8467 | |
| VCDModel=LLaVA-v1.52024.11 | 0.8185 | 0.8282 | |
| ICT w/o objectModel=LLaVA-v1.52024.11 | 0.8165 | 0.8132 | |
| RegularModel=LLaVA-v1.52024.11 | 0.799 | 0.7959 |