🧪 Test?View on arXiv
LWCal: Loss-Weighted Calibration for Tabular Classifiers with Noisy Calibration Labels
Author1, Author2, Author3, Author4, Author5
calibrationtabular datamachine learningprobability estimation
2609.26839
Builder Relevance
2h ago80%
Abstract
LWCal addresses the challenge of noisy calibration labels in post-hoc probability calibration for tabular classifiers.
Reality Card
Core Claim
LWCal achieves the lowest average calibration error on nine local binary tabular tasks without requiring clean validation labels or retraining.
Method / Result
Gated-LWCal reduces expected calibration error from 0.188 to 0.122.
Limitations
The method's performance may vary with different noise rates and types of label corruption, which could affect reproducibility.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.