Cross-Question Method Reuse in Large Language Models: From Word-Level Prediction to Rational Logical-Layer Reasoning
Hong Su
TL;DR
This work extends method-based reasoning in LLMs beyond highly similar questions by introducing cross-question method reuse. It formalizes two reuse modalities—relationship-based reuse based on general–specific or parallel relationships, and feature-based reuse leveraging partial or latent feature overlaps—as well as global methods and a hierarchical method-of-methods framework. The authors provide a theoretical account of reasoning at the logical layer and demonstrate rational reuse, ensuring solutions are applied only when transferable and valid. Verification experiments show that relationship-based reuse yields stronger, more reliable improvements than feature-based reuse, while feature-based reuse offers complementary coverage for less explicit similarities, together enhancing the robustness and versatility of LLM-driven problem solving.
Abstract
Large language models (LLMs) have been widely applied to assist in finding solutions for diverse questions. Prior work has proposed representing a method as a pair of a question and its corresponding solution, enabling method reuse. However, existing approaches typically require the questions to be highly similar. In this paper, we extend the scope of method reuse to address questions with low similarity or with hidden similarities that are not explicitly observable. For questions that are similar in a general-specific sense (i.e., broader or narrower in scope), we propose to first separate the question and solution, rather than directly feeding the pair to the LLM. The LLM is then guided to adapt the solution to new but related questions, allowing it to focus on solution transfer rather than question recognition. Furthermore, we extend this approach to cases where questions only share partial features or hidden characteristics. This enables cross-question method reuse beyond conventional similarity constraints. Experimental verification shows that our scope-extension approach increases the probability of filtering out reusable solutions, thereby improving the effectiveness of cross-question method reuse.