Generalization Suite
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Generalization Suite Unseen: RESISC45, SVHN ViT-B/32 (Split 2)
86.7Accuracy (Cars)
8
Generalization Suite Unseen EuroSAT MNIST ViT-B/32 (Split 1)
86.3Cars Accuracy
8
Generalization Suite Motion-X Kungfu EMDB AIST++ Video-Convert (test)
60.6Kungfu Score
8
Generalization Suite AG News, SST-2, TREC, CSQA, PIQA, SST-5, MR, MRPC, CB, COPA, CREAK, AI2SciE
88.2Accuracy (AG News)
6