Multimodal Hallucination Evaluation on CRPE
96.8Existence ScoreIC-VCO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| IC-VCOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Contrastive Sample Editing2026.05 | 96.8 | 72.52 | |
| IC-VCOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Synthetic2026.05 | 95.94 | 71.75 | |
| SymMPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Synthetic2026.05 | 95.67 | 71.45 | |
| mDPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Contrastive Sample Editing2026.05 | 95.61 | 71.93 | |
| mDPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Synthetic2026.05 | 95.54 | 71.24 | |
| S-VCOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Contrastive Sample Editing2026.05 | 95.48 | 72.04 | |
| SymMPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Contrastive Sample Editing2026.05 | 95.48 | 71.83 | |
| V-DPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Contrastive Sample Editing2026.05 | 95.46 | 72.19 | |
| V-DPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Synthetic2026.05 | 95.45 | 71.71 | |
| DPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Synthetic2026.05 | 95.43 | 71.76 | |
| S-VCOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Synthetic2026.05 | 95.3 | 71.18 | |
| DPOBackbone=LLaVA-OneVision-Qwen2-7B, Contrastive Sample Source=Contrastive Sample Editing2026.05 | 94.94 | 71.53 | |
| LLaVA-OneVision-Qwen2-7BBackbone=LLaVA-OneVision-Qwen2-7B2026.05 | 94.52 | 65.26 |