CORE: A Cognitive Science-Inspired Framework for Task-Adaptive Reasoning Enhancement in Large Language Models

Key Points

• Improvement in reasoning enhancement is observed with a framework inspired by cognitive science.
• The framework successfully adapts reasoning processes for large language models across various tasks.
• Analysis utilized existing cognitive principles to inform enhancements in language model performance.
• May enable better performance in AI applications requiring dynamic reasoning adjustments.