Introduces Relative Geometric Conflict (RGC) using gradient direction comparison and empirical Fisher trace factorization to improve reliability estimation beyond loss or confidence signals in noisy-label training.
L2b: Learning to bootstrap robust models for combating label noise
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
-
Radial-Angular Geometry for Reliable Update Diagnosis in Noisy-Label Learning
Introduces Relative Geometric Conflict (RGC) using gradient direction comparison and empirical Fisher trace factorization to improve reliability estimation beyond loss or confidence signals in noisy-label training.