A method for Model Fine Tuning that trains it off a small labelled dataset to adjust its weights better to a specific task. Best used for sentiment analysis and text classification.