A response set is the collection of outputs produced by a large language model for a given prompt set. It is used in evaluation workflows to compare quality, safety, consistency, or other model behavior across multiple prompts.
The collection of responses a large language model returns to an input prompt set.