LLM Learning Hub

workspace/llm-course/home

Overfitting

Overfitting occurs when a model memorizes training data patterns — including noise — and fails to generalize to unseen data. It shows up as low training loss but high validation loss.

What is Overfitting?

Overfitting occurs when a model memorizes training data patterns — including noise — and fails to generalize to unseen data. It shows up as low training loss but high validation loss.

Overfitting occurs when a model memorizes training data patterns — including noise — and fails to generalize to unseen data. It shows up as low training loss but high validation loss.

Where is it used?

Small LLM fine-tunes easily overfit narrow datasets; without regularization, deep models can memorize entire training sequences verbatim, hurting downstream evals.

How to build it

Train a high-capacity MLP on a tiny dataset, plot train vs. validation loss, and watch the gap widen — then add dropout and weight decay to close it.