LGM: Enhancing Large Language Models with Conceptual Meta-Relations and Iterative Retrieval
This addresses the issue of conceptual ambiguity in LLMs for users requiring accurate interpretations, though it appears incremental as it builds on RAG methods with specific enhancements.
The paper tackles the problem of large language models struggling with ambiguous or conceptually misaligned terms in user instructions by proposing the Language Graph Model (LGM), which extracts meta-relations like inheritance, alias, and composition and uses iterative retrieval to enhance conceptual clarity, resulting in consistent outperformance over existing RAG baselines on standard benchmarks.
Large language models (LLMs) exhibit strong semantic understanding, yet struggle when user instructions involve ambiguous or conceptually misaligned terms. We propose the Language Graph Model (LGM) to enhance conceptual clarity by extracting meta-relations-inheritance, alias, and composition-from natural language. The model further employs a reflection mechanism to validate these meta-relations. Leveraging a Concept Iterative Retrieval Algorithm, these relations and related descriptions are dynamically supplied to the LLM, improving its ability to interpret concepts and generate accurate responses. Unlike conventional Retrieval-Augmented Generation (RAG) approaches that rely on extended context windows, our method enables large language models to process texts of any length without the need for truncation. Experiments on standard benchmarks demonstrate that the LGM consistently outperforms existing RAG baselines.