Conceptual Metaphor Theory as a Prompting Paradigm for Large Language Models

2025-02-04

AI Summary

Large language models (LLMs) exhibit insufficient reasoning depth and ambiguous explanations in complex abstract reasoning and metaphor comprehension tasks. Method: This work introduces the first systematic computational translation of Conceptual Metaphor Theory (CMT) from cognitive linguistics into a prompt engineering paradigm. We propose a metaphor-mapping–based structured prompt template that explicitly embeds source-domain–target-domain mappings into the reasoning chain, guiding LLMs to emulate human-like metaphorical reasoning. We fine-tune multiple models—including Llama3.2, Phi3, Gemma2, and Mistral—and develop an automated evaluation framework built on Llama3.3-70B. Contribution/Results: CMT-enhanced models achieve significant improvements across diverse benchmarks: +12.4% in reasoning accuracy, +18.7% in explanation clarity, and enhanced metaphorical logical coherence—consistently outperforming baselines. This approach advances LLM interpretability and cognitive alignment through theoretically grounded, computationally operationalized linguistic principles.

Technology Category

• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

We introduce Conceptual Metaphor Theory (CMT) as a framework for enhancing large language models (LLMs) through cognitive prompting in complex reasoning tasks. CMT leverages metaphorical mappings to structure abstract reasoning, improving models' ability to process and explain intricate concepts. By incorporating CMT-based prompts, we guide LLMs toward more structured and human-like reasoning patterns. To evaluate this approach, we compare four native models (Llama3.2, Phi3, Gemma2, and Mistral) against their CMT-augmented counterparts on benchmark tasks spanning domain-specific reasoning, creative insight, and metaphor interpretation. Responses were automatically evaluated using the Llama3.3 70B model. Experimental results indicate that CMT prompting significantly enhances reasoning accuracy, clarity, and metaphorical coherence, outperforming baseline models across all evaluated tasks.

Problem

• Enhance large language models' reasoning.
• Improve abstract concept processing via metaphors.
• Compare models with CMT-augmented performance.

Innovation

• CMT enhances LLMs
• Metaphorical mappings structure reasoning
• CMT prompts improve reasoning accuracy