Fishr (ResNet-50)

Reported on 6 benchmarks across 2 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Methodology3 results

Domain AdaptationonOffice-Home
Average Accuracy· 2021-09-07
68.2
best: 90.6 (MoA (OpenCLIP, ViT-B/16))
Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization arXiv:2109.02934
Domain AdaptationonDomainNet
Average Accuracy· 2021-09-07
41.8
best: 67.4 (L2C (CLIP, ViT-L/14))
Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization arXiv:2109.02934
Domain AdaptationonVLCS
Average Accuracy· 2021-09-07
78.2
best: 85.5 (CAR-FT (CLIP, ViT-B/16))
Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization arXiv:2109.02934

Computer Vision3 results

Domain GeneralizationonOffice-Home
Average Accuracy· 2021-09-07
68.2
best: 90.6 (MoA (OpenCLIP, ViT-B/16))
Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization arXiv:2109.02934
Domain GeneralizationonDomainNet
Average Accuracy· 2021-09-07
41.8
best: 67.4 (L2C (CLIP, ViT-L/14))
Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization arXiv:2109.02934
Domain GeneralizationonVLCS
Average Accuracy· 2021-09-07
78.2
best: 85.5 (CAR-FT (CLIP, ViT-B/16))
Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization arXiv:2109.02934