What does Label Smoothing mean? Definition | AI Understanding

Definition

A regularization method that softens hard labels to improve generalization.

Related terms

Data Augmentation

Techniques 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.

Human-added labels or metadata used to train or evaluate machine learning models.

A simple reference model used to compare whether more complex approaches actually improve results.

Combining predictions from multiple models to improve robustness or accuracy.

How 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

See also

Knowledge Graph Latency Knowledge Distillation Large Language Model (LLM)Knowledge Cutoff Learning Rate Jailbreak Intent Classification

Browse the full AI Glossary Explore all free AI guides