Reasoning on Graphs: Faithful and Interpretable Large Language Model Reasoning

预备知识

核心思想

20260330174144

注: 其实, RoG 本质上就是认为构造了一些轨迹, 用 next-token prediction 训练而已.

参考文献

  1. Luo L., Li Y., Haffari G. and Pan S. Reasoning on Graphs: Faithful and Interpretable Large Language Model Reasoning. ICLR, 2024. [PDF] [Code]