AI Glossary Term

What does “Label Smoothing” mean?

Definition

A regularization method that softens hard labels to improve generalization.

Browse the full AI Glossary Explore all free AI guides

How Label Smoothing works

Example

Instead of telling a classifier the right class has probability 1 and every other class 0, training uses 0.9 and spreads 0.1 among the others. This stops the model from becoming extremely overconfident.

Why it matters

Smoothing is a simple regularizer that can improve generalization and calibration. It is a small change with often useful effects.

Common questions

Does it always improve accuracy?

Not always, but it often helps slightly, especially in large classifiers.

Does it change what the model predicts?

It changes training targets, so the final probabilities are softer and the model is less extreme about its confidence.

Are there downsides?

It can reduce the usefulness of the model's outputs for distillation in some cases, because softened probabilities carry less information.

Data AugmentationTechniques that create modified training examples to improve model generalization.LoRA (Low-Rank Adaptation)A parameter-efficient fine-tuning method that adds low-rank adapter matrices.AnnotationHuman-added labels or metadata used to train or evaluate machine learning models.Baseline ModelA simple reference model used to compare whether more complex approaches actually improve results.EnsembleCombining predictions from multiple models to improve robustness or accuracy.GeneralizationHow well a model performs on new, unseen data outside the training set.

Learn more in our free guides

Pseudo-Labeling and Self-Training Label Smoothing Semantic Role Labeling China's Deep Synthesis Rules and AI Content Labeling

Test yourself

Pseudo-Labeling and Self-Training Quiz Label Smoothing Quiz Semantic Role Labeling Quiz