Fine-tuning

Additional training on a smaller dataset after initial training, intended to adjust behavior, style, or domain coverage. Distinct from retrieval and from one-off prompting at inference time.