Browse Teach Curate

Graph-Enhanced Chain-of-Thought Reasoning for Academic Question Answering

A method for grounding a large language model's chain-of-thought reasoning in a heterogeneous scholarly graph, so that answers to academic questions are both structurally informed and interpretable. It converts graph structure into reasoning chains the model can process and adaptively selects query-relevant metapaths and subgraphs to supply as reasoning context.

Data Text

HetGCoT: Heterogeneous Graph-Enhanced Chain-of-Thought LLM

Primary NLP research (ACL EMNLP 2025 Findings) proposing HetGCoT, a framework that lets a large language model reason over a heterogeneous scholarly graph to answer academic questions interpretably. …