Proposes memorization-guided two-stage scoring to select debiased training subsets, enabling ERM models to achieve better performance than SOTA debiasing techniques using only 10% of data.
Avoiding spurious correlations via logit correction.arXiv preprint arXiv:2212.01433, 2022
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Mitigating Spurious Correlations with Memorization-Guided Dataset De-Biasing
Proposes memorization-guided two-stage scoring to select debiased training subsets, enabling ERM models to achieve better performance than SOTA debiasing techniques using only 10% of data.