Concretized Proposition Prompting Resolves Composition-Knowledge Dichotomy in Large Language Models
For LLM users and researchers, CPP offers a practical prompt framework that improves both factual accuracy and logical reasoning, addressing a known trade-off.
LLMs struggle with balancing compositionality and knowledgeability. The authors propose Concretized Proposition Prompting (CPP), which significantly enhances reasoning performance, especially in medical benchmarks where precise knowledge is critical, while remaining competitive on math benchmarks.
LLMs often struggle to balance compositionality with knowledgeability, a challenge we define as Composition-Knowledge Dichotomy. To address this, we propose Concretized Proposition Prompting (CPP), a framework that explicitly concretizes propositions relevant to questions. The results demonstrate that CPP significantly enhances reasoning performance, particularly in medical benchmarks where precise knowledge is paramount, while being competitive on math benchmarks where deductive reasoning is prioritized. Additional experiments reveal that CPP is scalable to various foundation models and parameter sizes, being a fundamental paradigm that bridges the gap between composition- and knowledge-based approaches. Consequently, CPP resolves the composition-knowledge dichotomy by providing a solid foundation for logically organized and factually grounded reasoning.