Learning with noisy labels in multimedia classification often combines external annotations and model predictions into a single reliability weight, even though the two sources can fail for different reasons. We instead estimate disentangled reliabilities: bilevel meta-learning produces two batch-normalized scalars per sample, alpha for the given label and beta for the pseudo-label, without constraining them to sum to one. Holistic Reliability Propagation (HRP) then routes them to different objectives, using reliability-aware Mixup with global gating on the input branch and beta-gated pseudo-label positives on the contrastive branch. On synthetic and real-world benchmarks, HRP improves average accuracy over strong baselines and remains competitive at the highest noise rates.
@article{arxiv.2605.20725,
title = {Holistic Reliability Propagation: Decoupling Annotation and Prediction for Robust Noisy-Label},
author = {Jingyang Mao and Ningkang Peng and Yanhui Gu},
journal= {arXiv preprint arXiv:2605.20725},
year = {2026}
}