Title: uTeBC-NLP at SemEval-2024 Task 9: Can LLMs be Lateral Thinkers?

This research evaluates LLMs’ potential for lateral thinking, the cognitive ability to solve problems through an indirect and creative approach. It experiments with various prompting techniques to enhance this ability in LLMs, such as chain of thoughts and direct prompts. The study employs models like GPT-3.5 and GPT-4, and reveals that dynamic in-context learning significantly improves lateral thinking in NLP tasks.

• Explores prompting methods to boost LLM’s lateral thinking in language tasks.
• Utilizes advanced models like GPT-3.5 and GPT-4 for experimentation.
• Highlights performance enhancement through condensed and dynamic prompts.
• Provides insights into the potential of LLMs for innovative problem-solving.

My Opinion: Lateral thinking is a nuanced cognitive process, and its emulation by LLMs would be a monumental stride towards AI understanding complex human thought processes. The study’s findings on prompt efficiency are especially intriguing, suggesting avenues for refining LLMs’ creative cognition.