Visual Personalization on Visual Personalization Evaluation Set
97.4VIPER Proxy Score (PS)VPRAG
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VPRAGComparison Baseline=GPT-image-1, Sample Budget=@1002026.01 | 97.4 | 44 | 3.73 | 33 | |
| GPT-image-1Sample Budget=@1002026.01 | 96.6 | — | 3.86 | — | |
| VPRAGComparison Baseline=Flux DB-LoRA, Sample Budget=@502026.01 | 86.7 | 80 | 3.17 | 88 | |
| VPRAGComparison Baseline=GPT (VLM + GPT-image-1), Sample Budget=@1002026.01 | 85.8 | 63 | 3.28 | 49 | |
| VPRAGComparison Baseline=DrUM, Sample Budget=@10002026.01 | 84.1 | 68.9 | 3.41 | 76.4 | |
| VPRAGComparison Baseline=Flux Kontext, Sample Budget=@10002026.01 | 83.9 | 83.4 | 3.11 | 76.4 | |
| GPT (VLM + GPT-image-1)Sample Budget=@1002026.01 | 83.2 | — | 3.22 | — | |
| DrUMSample Budget=@10002026.01 | 75.7 | — | 2.89 | — | |
| VPRAGComparison Baseline=Viper, Sample Budget=@10002026.01 | 67.8 | 71.5 | 2.88 | 61.8 | |
| Flux DB-LoRASample Budget=@502026.01 | 65.6 | — | 1.99 | — | |
| ViperSample Budget=@10002026.01 | 54.5 | — | 2.4 | — | |
| Flux KontextSample Budget=@10002026.01 | 54.1 | — | 2.29 | — |