Parameter-Efficient Fine-Tuning

Parameter-efficient fine-tuning is a family of techniques (LoRA, adapters, prefix tuning, prompt tuning) that adapt a large pretrained model to a new task or domain by updating only a small fraction of its parameters - typically under 1% - while freezing the rest. This dramatically reduces memory, storage, and compute cost compared with full fine-tuning, making it feasible to specialise foundation models for niche accessibility tasks (sign language recognition, dysarthric speech, clinical dialogue) on modest hardware without sacrificing base-model capability.