Classification on TAG-C corrupted visual modality v1 (test)
61.6Brightness Accuracy (TAG-C)RobustTouch
Evaluation Results
| Method | Links | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RobustTouchEvaluation Scenario=continuous cross-domain setting, Corruption Severity=3, Optimizer=AdamW, Learning Rate=1e-62026.01 | 61.6 | 61.7 | 65 | 51.2 | 62.3 | 64.8 | 66.7 | 68.6 | 67.7 | 70.5 | 55.6 | 57.7 | 60.5 | 62.4 | 53 | 57.7 | |
| No Adapt.Evaluation Scenario=continuous cross-domain setting, Corruption Severity=3, Backbone=CLIP (Anytouch)2026.01 | 61 | 57.9 | 61.3 | 50.2 | 58.9 | 60.6 | 59.4 | 60.9 | 57.2 | 62.6 | 48.8 | 57.2 | 62.2 | 62.8 | 49.2 | 55.7 | |
| SAREvaluation Scenario=continuous cross-domain setting, Corruption Severity=3, Mode=Single-Modal TTA2026.01 | 61 | 57.8 | 61.3 | 50.2 | 58.9 | 60.6 | 59.4 | 60.8 | 57.1 | 62.6 | 48.9 | 57.3 | 62.2 | 62.8 | 49.2 | 55.5 | |
| TDAEvaluation Scenario=continuous cross-domain setting, Corruption Severity=3, Mode=Multi-Modal TTA2026.01 | 43.1 | 32.1 | 41.8 | 21.2 | 30.2 | 34.9 | 40.2 | 36.7 | 27.4 | 35 | 15.5 | 33.5 | 36.1 | 38.9 | 17.8 | 29.2 | |
| READEvaluation Scenario=continuous cross-domain setting, Corruption Severity=3, Mode=Multi-Modal TTA2026.01 | 34.5 | 11.8 | 13.4 | 11.3 | 10.6 | 10.2 | 10.1 | 9.8 | 10 | 9.8 | 10 | 9.5 | 9.7 | 9.7 | 9.3 | 9.3 | |
| TENTEvaluation Scenario=continuous cross-domain setting, Corruption Severity=3, Mode=Single-Modal TTA2026.01 | 31.7 | 23.7 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 | 23.1 |