Prompt Engineering Evaluation Metrics for Interpretable Joint Evaluation of Prompts and Responses

• Propose a Prompt Engineering Evaluation Metrics (PEEM), a novel framework designed to assess the prompt's linguistic quality and task alignment.

• To validate PEEM's evaluation capabilities, we applied PEEM's result to guide prompt rewriting.

• To further challenge our method, we compared existing prompt optimization methods with PEEM-based rewriting.

• We conducted experiments to demonstrate that our framework achieves comparable performance with reinforcement learning-based fine-tuning methods, even within zero-shot settings.

Outcome

• IEEE Access / Under-Review