Free-form Captioning Hallucination Evaluation on CHAIR
59.6CHAIR-S ScoreVCD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VCDModel=LLaVA-1.5, Max Token=1282026.06 | 59.6 | 16.6 | |
| PAIModel=InstructBLIP, Max Token=1282026.06 | 56.8 | 16.7 | |
| RegularModel=LLaVA-1.5, Max Token=1282026.06 | 56.6 | 16.8 | |
| VCDModel=InstructBLIP, Max Token=1282026.06 | 55.8 | 15.9 | |
| ONLYModel=InstructBLIP, Max Token=1282026.06 | 55.8 | 17.4 | |
| RegularModel=InstructBLIP, Max Token=1282026.06 | 54.2 | 16.7 | |
| PAIModel=LLaVA-1.5, Max Token=1282026.06 | 54 | 15.2 | |
| CAIModel=InstructBLIP, Max Token=1282026.06 | 52.8 | 16.5 | |
| ONLYModel=LLaVA-1.5, Max Token=1282026.06 | 48.4 | 14.7 | |
| CAIModel=LLaVA-1.5, Max Token=1282026.06 | 39.6 | 12.4 | |
| RegularModel=InstructBLIP, Max Token=642026.06 | 29 | 10.2 | |
| VCDModel=InstructBLIP, Max Token=642026.06 | 26.4 | 8.7 | |
| RegularModel=LLaVA-1.5, Max Token=642026.06 | 26 | 8.8 | |
| PAIModel=LLaVA-1.5, Max Token=642026.06 | 26 | 8.7 | |
| ONLYModel=InstructBLIP, Max Token=642026.06 | 25.2 | 8.8 | |
| PAIModel=InstructBLIP, Max Token=642026.06 | 24.6 | 8.3 | |
| CAIModel=InstructBLIP, Max Token=642026.06 | 24.2 | 8.2 | |
| VCDModel=LLaVA-1.5, Max Token=642026.06 | 23.8 | 8.2 | |
| ONLYModel=LLaVA-1.5, Max Token=642026.06 | 21 | 7.3 | |
| CAIModel=LLaVA-1.5, Max Token=642026.06 | 17.8 | 6.9 |