Showing cs.LGShow all
3 papers · 1 filter
cs.LG2026
Corroborative Learning: Mutual Enhancement through Task-Response Reconstruction
Zexing Zhang, Jichao Li, Tianyang Lei +2
In pattern analysis, models trained independently for the same task often develop complementary capabilities through differences in architecture, data, and optimization. Exploiting…
cs.LG2026
Beyond Scarcity: When Encoder Adaptation Is Worthwhile
Zexing Zhang, Jichao Li, Tianyang Lei +2
Long-tailed recognition often learns a shared representation before correcting the classifier for class imbalance. Once the classifier has been rebalanced, however, final accuracy…
cs.LG2026
Selective Posterior Margin Regularization for Forward-Corrected Classification
Zexing Zhang, Jichao Li, Tianyang Lei +2
Learning with class-conditional label noise often relies on a transition model from latent clean classes to observed annotations. Forward correction embeds this transition in the l…