Creativity, often thought to be a uniquely human trait, is being assessed in artificial intelligence. A recent study adapts the Torrance Tests of Creative Thinking to evaluate the performance of various large language models (LLMs) on tasks measuring Fluency, Flexibility, Originality, and Elaboration.

• The study introduces a dataset of 700 questions aimed at testing LLMs’ creativity.
• Findings suggest LLMs are exceptional in Elaboration but fall short in Originality.
• Prompts and role-play strongly influence LLMs’ creative outputs.
• Collaboration between multiple LLMs may enhance creativity, particularly originality.
• There’s an observed consensus between human and LLM evaluations on personality traits affecting creativity.

Understanding the creative capabilities of LLMs is fundamental for advancements in NLP. This paper provides a structured framework for assessing and enhancing creativity in AI, potentially leading to more genuine and innovative content generation.