LLM reasoning enhanced via external subgraph generation

Arxiv CS.CL (NLP/language models)

The method generates external subgraphs to improve stepwise reasoning in large language models. It targets logical consistency, factual grounding, and interpretability in complex multi-step tasks.

llmreasoning

Arxiv CS.CL (NLP/language models)·Xin ZhangYang CaoBaoxing WuKai SongSiying Li

7 days ago