A family of methods (like LoRA) that adapt large models by training only a tiny fraction of parameters.
← All terms