SST uses per-class thresholds updated once per cycle to pick pseudo-labels, reporting 84.9% ImageNet top-1 accuracy with 10% labeled data on a huge ViT.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SST: Self-training with Self-adaptive Thresholding for Semi-supervised Learning
SST uses per-class thresholds updated once per cycle to pick pseudo-labels, reporting 84.9% ImageNet top-1 accuracy with 10% labeled data on a huge ViT.