Guardado en:
Detalles Bibliográficos
Autores principales: Singh, Agamdeep, Tiwari, Ashish, Hasanbeig, Hosein, Gupta, Priyanshu
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2511.14117
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911273701081088
author Singh, Agamdeep
Tiwari, Ashish
Hasanbeig, Hosein
Gupta, Priyanshu
author_facet Singh, Agamdeep
Tiwari, Ashish
Hasanbeig, Hosein
Gupta, Priyanshu
contents Many machine learning tasks involve inherent subjectivity, where annotators naturally provide varied labels. Standard practice collapses these label distributions into single labels, aggregating diverse human judgments into point estimates. We argue that this approach is epistemically misaligned for ambiguous data--the annotation distribution itself should be regarded as the ground truth. Training on collapsed single labels forces models to express false confidence on fundamentally ambiguous cases, creating a misalignment between model certainty and the diversity of human perception. We demonstrate empirically that soft-label training, which treats annotation distributions as ground truth, preserves epistemic uncertainty. Across both vision and NLP tasks, soft-label training achieves 32% lower KL divergence from human annotations and 61% stronger correlation between model and annotation entropy, while matching the accuracy of hard-label training. Our work repositions annotation distributions from noisy signals to be aggregated away, to faithful representations of epistemic uncertainty that models should learn to reproduce.
format Preprint
id arxiv_https___arxiv_org_abs_2511_14117
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Soft-Label Training Preserves Epistemic Uncertainty
Singh, Agamdeep
Tiwari, Ashish
Hasanbeig, Hosein
Gupta, Priyanshu
Machine Learning
Artificial Intelligence
Many machine learning tasks involve inherent subjectivity, where annotators naturally provide varied labels. Standard practice collapses these label distributions into single labels, aggregating diverse human judgments into point estimates. We argue that this approach is epistemically misaligned for ambiguous data--the annotation distribution itself should be regarded as the ground truth. Training on collapsed single labels forces models to express false confidence on fundamentally ambiguous cases, creating a misalignment between model certainty and the diversity of human perception. We demonstrate empirically that soft-label training, which treats annotation distributions as ground truth, preserves epistemic uncertainty. Across both vision and NLP tasks, soft-label training achieves 32% lower KL divergence from human annotations and 61% stronger correlation between model and annotation entropy, while matching the accuracy of hard-label training. Our work repositions annotation distributions from noisy signals to be aggregated away, to faithful representations of epistemic uncertainty that models should learn to reproduce.
title Soft-Label Training Preserves Epistemic Uncertainty
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2511.14117