Fine-tuning

A machine-learning technique that adapts a pre-trained foundation model - typically a large language model or vision model - to a specific task, domain, or individual user by continuing training on a smaller, targeted dataset. Fine-tuning preserves the broad capabilities of the base model while shifting its outputs toward a desired style, vocabulary, or behavior. In accessibility, fine-tuning is used to personalize speech-generating devices to an AAC user's own communication patterns, to adapt speech recognition to atypical speech (dysarthria, deafened speech), to tailor image-description models to a particular user's preferences, and to train assistive gesture or pose models on data that includes disabled users. Fine-tuning raises identity, privacy, and bias considerations because the resulting model carries traces of the training data into every future output.