Hallucination Evaluation on POPE COCO 2014 (Adversarial)
81.43AccuracyVEGAS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VEGASBase Model=LLaVA-1.5, Decoding Strategy=Greedy2025.12 | 81.43 | 81.78 | |
| VISTABase Model=LLaVA-1.5, Decoding Strategy=Greedy2025.12 | 81.3 | 82.7 | |
| PAIBase Model=LLaVA-1.5, Decoding Strategy=Greedy2025.12 | 81.05 | 82.17 | |
| VanillaBase Model=LLaVA-1.5, Decoding Strategy=Greedy2025.12 | 80.03 | 81.22 | |
| VCDBase Model=LLaVA-1.5, Decoding Strategy=Greedy2025.12 | 79.95 | 81.17 | |
| VEGASBase Model=Shikra, Decoding Strategy=Greedy2025.12 | 78.94 | 79.95 | |
| VanillaBase Model=Shikra, Decoding Strategy=Greedy2025.12 | 78.68 | 79.75 | |
| PAIBase Model=Shikra, Decoding Strategy=Greedy2025.12 | 78.68 | 79.78 | |
| VISTABase Model=Shikra, Decoding Strategy=Greedy2025.12 | 78.63 | 79 | |
| VEGASBase Model=MiniGPT-4, Decoding Strategy=Greedy2025.12 | 72.77 | 73.03 | |
| VISTABase Model=MiniGPT-4, Decoding Strategy=Greedy2025.12 | 72.47 | 72.78 | |
| PAIBase Model=MiniGPT-4, Decoding Strategy=Greedy2025.12 | 71.7 | 72.39 | |
| VanillaBase Model=MiniGPT-4, Decoding Strategy=Greedy2025.12 | 71.13 | 71.96 |