use case · basics
The simplest thing the model does: read a piece of text and decide whether it sounds positive or negative, with a confidence. Type in the box and the verdict updates a moment after you pause. Try flipping a single word — "love" to "hate" — and watch the meter swing.
Because it was trained on movie reviews, the model is confident on clear opinions and wobblier on neutral, factual, or heavily sarcastic text. That's a feature of the training data, not a bug — and a good reminder to check confidence, not just the label.
← Back to DistilBERT · Next: Practical →