Prompt Evaluation/Accuracy Assessment

Testing the Digital Human for Accuracy

The Digital Human is powered by OpenAI's Large Language Model (LLM), which utilizes a transformer-based neural network architecture designed to process and generate text.

1. Create 20 Questions: Develop a set of 20 clear and specific questions based on the content of the document.

2. Test the AI: Use your questions to evaluate the AI’s responses, comparing them against the expected answers from the document.

3. Review and Improve: Identify any issues, such as knowledge gaps, repetition, or conflicting information, and update the content as needed.

4. Repeat (if needed): Continue testing and refining until the AI’s responses consistently meet your expectations.

5. Submit to UNITH for Scoring: Send the 20 questions to UNITH at [email protected] for an automated accuracy score..