Visual consistency assessment on VCReward-Bench
93.59Subject Addition ConsistencyGemini 3 Pro
Evaluation results
| Method | Links | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Gemini 3 Pro2026.03 | 93.59 | 86.32 | 85.88 | 83.52 | 92.9 | 86.91 | 81.42 | 85.06 | 85.94 | 87.83 | 96.72 | 94.66 | 87.07 | 84.08 | 86.6 | 85.05 | 88.28 | 76.27 | 81.01 | 89.42 | 91.18 | 87.13 | |
| PVC-Judge2026.03 | 86.96 | 74.19 | 82.57 | 72.35 | 87.04 | 86.84 | 75.69 | 84.32 | 82.78 | 78.21 | 85.29 | 89.39 | 86.13 | 84.24 | 86.87 | 80.17 | 79.69 | 66.12 | 80.22 | 81.12 | 88.07 | 81.82 | |
| GPT 5.12026.03 | 80.68 | 68.49 | 67.96 | 69.45 | 75.96 | 80.09 | 70.99 | 79.89 | 72.97 | 79.33 | 83.97 | 92.25 | 80.08 | 82.91 | 85.57 | 72.41 | 82.54 | 68.85 | 65.68 | 77.86 | 76.74 | 76.89 | |
| Qwen3-VL-8B-Instruct2026.03 | 72.53 | 69.01 | 74 | 68.22 | 67.01 | 72.89 | 76.23 | 72.93 | 71.33 | 77.87 | 83.09 | 87.6 | 72.69 | 73.62 | 78.79 | 69.83 | 76.56 | 51.37 | 68.68 | 68.97 | 81.31 | 73.07 | |
| EditReward2026.03 | 67.39 | 70.97 | 65.14 | 57.34 | 75.46 | 57.02 | 59.03 | 57.84 | 81.46 | 76.92 | 90.44 | 92.42 | 53.36 | 56.36 | 49.49 | 68.97 | 60.94 | 56.83 | 63.19 | 67.86 | 87.16 | 67.41 | |
| EditScore Avg@42026.03 | 41.3 | 35.48 | 36.7 | 34.81 | 53.24 | 50.88 | 34.03 | 37.84 | 50.99 | 50 | 58.82 | 82.58 | 50.84 | 41.21 | 65.66 | 57.76 | 57.03 | 31.69 | 48.35 | 56.12 | 57.8 | 49.2 |