Cross-distribution performance on SAIR official benchmark (hard3)
81.3Accuracy
hard3_overfitted
Updated 1mo ago
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
2026.04 | 81.3 | 83.1 | 30.3 | |
2026.04 | 71 | 76 | 11 | |
2026.04 | 70.8 | 74.6 | 12.2 | |
2026.04 | 68 | 74.3 | 18.5 | |
2026.04 | 65.3 | 67.6 | 24.3 | |
2026.04 | 61.3 | 68.6 | 27.7 | |
2026.04 | 60.8 | 69.3 | 9.3 | |
2026.04 | 59 | 62.6 | 21 | |
2026.04 | 58.5 | 66.5 | 27.5 | |
2026.04 | 58.3 | 67.3 | 0.3 | |
2026.04 | 56.3 | 55.9 | 42.7 | |
2026.04 | 55.5 | 53.9 | 41.5 | |
2026.04 | 55.5 | 63.7 | 7.5 |