Hallucination Evaluation on POPE MSCOCO, A-OKVQA, GQA average (Adversarial)
84.11AccuracyVanilla
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VanillaMLLM=Qwen3-VL2026.02 | 84.11 | 80.25 | 91.35 | 85.24 | |
| AdaVBoostMLLM=Qwen3-VL2026.02 | 84.09 | 80.38 | 91.04 | 85.17 | |
| PAIMLLM=Qwen3-VL2026.02 | 83.55 | 78.96 | 92.6 | 85 | |
| VGAMLLM=Qwen3-VL2026.02 | 83.05 | 78.85 | 91.29 | 84.4 | |
| VAFMLLM=Qwen3-VL2026.02 | 82.39 | 77.17 | 93.11 | 84.17 | |
| VanillaMLLM=LLaVA-NeXT2026.02 | 80.18 | 74.47 | 93.82 | 82.72 | |
| AdaVBoostMLLM=LLaVA-NeXT2026.02 | 80.15 | 74.46 | 93.8 | 82.7 | |
| VGAMLLM=InternVL3.52026.02 | 78.71 | 72.04 | 94.5 | 81.66 | |
| VanillaMLLM=InternVL3.52026.02 | 78.68 | 72.71 | 92.59 | 81.3 | |
| VAFMLLM=InternVL3.52026.02 | 78.48 | 71.55 | 95.17 | 81.61 | |
| AdaVBoostMLLM=InternVL3.52026.02 | 78.46 | 71.61 | 94.91 | 81.54 | |
| VAFMLLM=LLaVA-NeXT2026.02 | 77.81 | 71.01 | 95.89 | 81.36 | |
| VGAMLLM=LLaVA-NeXT2026.02 | 77.46 | 71.87 | 93.22 | 80.77 | |
| PAIMLLM=LLaVA-NeXT2026.02 | 77.23 | 70.95 | 95.13 | 80.93 | |
| PAIMLLM=InternVL3.52026.02 | 77.22 | 70.71 | 93.53 | 80.44 |