TMLC-Net trains an LSTM to output softened versions of the noisy labels, then claims this transfers as label correction to new datasets without ever seeing a true clean label.
Learning to learn from noisy labeled data,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
TMLC-Net: Transferable Meta Label Correction for Noisy Label Learning
TMLC-Net trains an LSTM to output softened versions of the noisy labels, then claims this transfers as label correction to new datasets without ever seeing a true clean label.