The rapid evolution of Large Language Models (LLMs) has sparked widespread interest in their creative potential. A recent study dives into the methods to assess such creativity, arguing that existing metrics are not sufficient and highlighting the need for a multi-dimensional measurement system.

• This paper adapts the modified Torrance Tests of Creative Thinking.
• It sets up an LLM-based method for evaluating creativity across multiple tasks, focusing on Fluency, Flexibility, Originality, and Elaboration.
• The study uses a dataset of 700 questions and evaluates various LLMs’ responses to diverse prompts.
• Results indicate a discrepancy in originality yet a proficiency in elaboration, revealing how design and prompt use significantly impact LLM creativity.

Opinion: Understanding the creativity of LLMs is vital for their application across various domains. This research showcases a structured approach to evaluate and improve LLMs, potentially aiding in the development of more innovative AI tools for creative tasks.