Image Attribution on VAR (External Fidelity Metrics)
100ImageNet FidelityOurs
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| OursOriginal Model=VAR2026.06 | 100 | 97.1 | 100 | 96.8 | 100 | 99.6 | 100 | 100 | |
| OursOriginal Model=VAR, Set of signals=best-performing2026.06 | 99.6 | 86.3 | 99.8 | 95.8 | 100 | 98.9 | 100 | 100 | |
| AEDROriginal Model=VAR2026.06 | 9.8 | 10.9 | 45.3 | 4.1 | 22.6 | 3.9 | 33.1 | 40.4 | |
| LatentTracerOriginal Model=VAR2026.06 | 2.8 | 0.1 | 8.4 | 0.1 | 4.2 | 0 | 12.8 | 10.1 | |
| ReconstructionOriginal Model=VAR2026.06 | 0.5 | 0.5 | 2.1 | 0.1 | 1.4 | 0 | 5.4 | 4 | |
| ReconstructionOriginal Model=VAR2026.06 | 0.3 | 0.2 | 0.4 | 0 | 0.5 | 0 | 1.9 | 0.3 | |
| LatentTracerOriginal Model=VAR2026.06 | 0.2 | 0 | 3.4 | 0 | 1.8 | 0 | 3.4 | 2.4 | |
| AEDROriginal Model=VAR2026.06 | 0 | 1.6 | 29.5 | 0.6 | 3.8 | 1.2 | 16.8 | 25.4 |