Classification Attribution on ImageNet misclassified samples (val)
0.4831Insertion ScoreGreedy
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GreedyBackbone=CLIP RN101, Attribution Target=Model's wrong prediction2026.06 | 0.4831 | 0.0309 | 0.744 | 0.6297 | 1,748.26 | 2.81 | |
| PhaseWinBackbone=CLIP RN101, Attribution Target=Model's wrong prediction2026.06 | 0.4264 | 0.0316 | 0.7175 | 0.5281 | 1,014.2 | 4.2 | |
| RISEBackbone=CLIP RN101, Attribution Target=Model's wrong prediction2026.06 | 0.2543 | 0.0602 | 0.5435 | 0.3443 | 5,000 | 1.27 | |
| HSICBackbone=CLIP RN101, Attribution Target=Model's wrong prediction2026.06 | 0.2365 | 0.0671 | 0.521 | 0.3292 | 1,536 | 4.73 | |
| Gradient IntegralBackbone=CLIP RN101, Attribution Target=Model's wrong prediction2026.06 | 0.193 | 0.115 | 0.5091 | 0.2251 | — | — | |
| GradientBackbone=CLIP RN101, Attribution Target=Model's wrong prediction2026.06 | 0.1891 | 0.112 | 0.5097 | 0.2168 | — | — | |
| IGOS++Backbone=CLIP RN101, Attribution Target=Model's wrong prediction2026.06 | 0.153 | 0.1127 | 0.5051 | 0.1367 | 49 | 31.22 |