Image Classification on TinyImageNet ε=1/255 (test)
43.78Standard AccuracyCC-DIST
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CC-DISTε=1/2552026.06 | 43.78 | 27.88 | |
| AD-CERTε=1/2552026.06 | 40.06 | 28.19 | |
| MTL-IBPε=1/2552026.06 | 35.97 | 27.73 | |
| TAPS/STAPSε=1/255, Selection protocol=Stronger of TAPS and STAPS based on certified accuracy2026.06 | 30.63 | 22.31 | |
| SABRε=1/2552026.06 | 30.58 | 20.96 | |
| IBPε=1/2552026.06 | 26.77 | 19.82 | |
| SORTNETε=1/255, Architecture=Non-ReLU-based2026.06 | 25.69 | 18.18 |