Object Hallucination on COCO captions 2014 (val)
12.3CHAIR (scene)Vanilla
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VanillaModel=Cham-7B2026.02 | 12.3 | 8.5 | 102.3 | |
| VanillaModel=Cham-30B2026.02 | 10.8 | 7.2 | 108.2 | |
| VanillaModel=LLaVA-1.5-7B2026.02 | 10.2 | 7.1 | 108.5 | |
| VCDModel=Cham-7B2026.02 | 9.8 | 6.8 | 97.4 | |
| OA-VCDModel=Cham-7B2026.02 | 9.5 | 6.3 | 99.2 | |
| CRoPS†Model=Cham-7B2026.02 | 9.1 | 5.8 | 98.2 | |
| VanillaModel=LLaVA-1.5-13B2026.02 | 9.1 | 6.3 | 114.8 | |
| VanillaModel=Qwen-VL2026.02 | 8.9 | 6.1 | 115.2 | |
| VCDModel=Cham-30B2026.02 | 8.5 | 5.5 | 103.8 | |
| OA-VCDModel=Cham-30B2026.02 | 8.2 | 5.6 | 105.2 | |
| VCDModel=LLaVA-1.5-7B2026.02 | 8.2 | 5.2 | 103.8 | |
| VanillaModel=Qwen2-VL-7B2026.02 | 8.2 | 5.7 | 118.5 | |
| CRoPS†Model=Cham-30B2026.02 | 7.9 | 5 | 104.4 | |
| OA-VCDModel=LLaVA-1.5-7B2026.02 | 7.9 | 5.3 | 105.3 | |
| VanillaModel=InternVL2-7B2026.02 | 7.8 | 5.4 | 121.4 | |
| CRoPS†Model=LLaVA-1.5-7B2026.02 | 7.6 | 5.3 | 104.5 | |
| VCDModel=LLaVA-1.5-13B2026.02 | 7.2 | 4.9 | 110.6 | |
| SCRModel=Cham-7B2026.02 | 7.1 | 4.8 | 101.7 | |
| VCDModel=Qwen-VL2026.02 | 7 | 4.7 | 111.2 | |
| OA-VCDModel=LLaVA-1.5-13B2026.02 | 6.9 | 4.4 | 112.2 | |
| OA-VCDModel=Qwen-VL2026.02 | 6.7 | 4.2 | 112.5 | |
| VCDModel=Qwen2-VL-7B2026.02 | 6.5 | 4.5 | 114.6 | |
| CRoPS†Model=Qwen-VL2026.02 | 6.4 | 4.5 | 111.8 | |
| VCDModel=InternVL2-7B2026.02 | 6.3 | 4.2 | 117.2 | |
| OA-VCDModel=Qwen2-VL-7B2026.02 | 6.2 | 4.1 | 115.4 | |
| SCRModel=LLaVA-1.5-7B2026.02 | 5.9 | 3.8 | 107.8 | |
| CRoPS†Model=Qwen2-VL-7B2026.02 | 5.9 | 3.9 | 115.1 | |
| OA-VCDModel=InternVL2-7B2026.02 | 5.9 | 3.9 | 118.5 | |
| SCRModel=Cham-30B2026.02 | 5.8 | 3.4 | 107.4 | |
| SCRModel=Qwen-VL2026.02 | 5.2 | 3.6 | 114.6 | |
| SCRModel=LLaVA-1.5-13B2026.02 | 4.7 | 3.3 | 114.1 | |
| SCRModel=InternVL2-7B2026.02 | 4.6 | 3 | 120.7 | |
| CRoPS†Model=LLaVA-1.5-13B2026.02 | 4.5 | 3 | 110.8 | |
| CRoPS†Model=InternVL2-7B2026.02 | 4.3 | 2.8 | 117.3 | |
| SCRModel=Qwen2-VL-7B2026.02 | 4 | 2.5 | 117.9 |