Image-to-Text Hallucination Evaluation on COCO
0.13CHAIRs ScoreBoN+CycRew
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BoN+CycRewBackbone=Qwen2.5-Omni-7B2026.05 | 0.13 | 0.082 | 0.598 | |
| WoodpeckerBackbone=Qwen2.5-Omni-7B2026.05 | 0.1 | 0.067 | 0.616 | |
| BoN+CLIPBackbone=Qwen2.5-Omni-7B2026.05 | 0.08 | 0.047 | 0.613 | |
| FrozenBackbone=Qwen2.5-Omni-7B2026.05 | 0.07 | 0.069 | 0.588 | |
| VolcanoBackbone=Qwen2.5-Omni-7B2026.05 | 0.07 | 0.042 | 0.637 | |
| BoN+VisPRMBackbone=Qwen2.5-Omni-7B2026.05 | 0.06 | 0.031 | 0.616 | |
| DeGFBackbone=Qwen2.5-Omni-7B2026.05 | 0.06 | 0.033 | 0.594 | |
| VCDBackbone=Qwen2.5-Omni-7B2026.05 | 0.06 | 0.045 | 0.61 | |
| FrozenBackbone=LLaVA-1.5-7B2026.05 | 0.06 | 0.028 | 0.74 | |
| BoN+CLIPBackbone=LLaVA-1.5-7B2026.05 | 0.06 | 0.025 | 0.746 | |
| TIGERBackbone=Qwen2.5-Omni-7B2026.05 | 0.05 | 0.035 | 0.643 | |
| BoN+VisPRMBackbone=LLaVA-1.5-7B2026.05 | 0.05 | 0.021 | 0.746 | |
| WoodpeckerBackbone=LLaVA-1.5-7B2026.05 | 0.05 | 0.025 | 0.74 | |
| VCDBackbone=LLaVA-1.5-7B2026.05 | 0.045 | 0.02 | 0.755 | |
| BoN+CycRewBackbone=LLaVA-1.5-7B2026.05 | 0.04 | 0.02 | 0.745 | |
| DeGFBackbone=LLaVA-1.5-7B2026.05 | 0.04 | 0.018 | 0.771 | |
| VolcanoBackbone=LLaVA-1.5-7B2026.05 | 0.04 | 0.023 | 0.753 | |
| TIGERBackbone=LLaVA-1.5-7B2026.05 | 0.03 | 0.013 | 0.772 |