Reasoning Ability Emerges in Large Language Models as Aggregation of Reasoning Paths: A Case Study With Knowledge Graphs

Xinyi Wang, William Yang Wang

Keywords: Large language models, reasoning, explanation

Abstract: This study focuses on the emergence of reasoning abilities in large language models (LLMs). While LLMs have shown remarkable capabilities in complex reasoning tasks, the exact origin of this ability and its relationship to pre-training and fine-tuning stages remain unclear. Previous research has explored in-context learning but has not fully addressed reasoning abilities such as logical reasoning or math deduction. The paper proposes investigating reasoning in LLMs through reasoning over knowledge graphs. The experiments demonstrate the importance of the pre-training sequence in enabling effective reasoning. The findings suggest that LLMs acquire reasoning abilities during pre-training rather than fine-tuning. Furthermore, training LLMs with next-token prediction enables them to aggregate relevant reasoning paths and derive new conclusions. The empirical results support the explanation of LLMs predicting unseen facts using a path ranking algorithm.

Submission Number: 12