Evaluation
Running a model against a fixed set of questions with a fixed expected answer, and scoring the result. An eval is how a change to a prompt is proven to be an improvement.
- See also: Hallucination
- See also: Regression
- Used in: LLMs
Running a model against a fixed set of questions with a fixed expected answer, and scoring the result. An eval is how a change to a prompt is proven to be an improvement.