Visual Question Answering on POPE Adversarial
85.86AccuracyLASER
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LASERModel=LLaVA-1.5, Decoding=LASER2026.02 | 85.86 | 85.56 | |
| ViCropModel=LLaVA-1.5, Decoding=ViCrop2026.02 | 85.8 | 84.93 | |
| VCDModel=LLaVA-1.5, Decoding=VCD2026.02 | 85.3 | 84.5 | |
| SampleModel=LLaVA-1.5, Decoding=Sample2026.02 | 83.7 | 82.23 | |
| LASERModel=Qwen-VL, Decoding=LASER2026.02 | 83.6 | 83.36 | |
| VCDModel=Qwen-VL, Decoding=VCD2026.02 | 83.15 | 82.79 | |
| ViCropModel=Qwen-VL, Decoding=ViCrop2026.02 | 82.57 | 82.28 | |
| SampleModel=Qwen-VL, Decoding=Sample2026.02 | 81.4 | 80.56 |