DoG

The offical codebase for Decoding on Graphs: Faithful and Sound Reasoning on Knowledge Graphs through Generation of Well-Formed Chains.

---

📌 Overview

This codebase is customized for LLaMA 3.1, providing a ready‑to‑run setup DoG. If you want to adapt it to other LLM series, you should modify relevant parts (mainly stop tokens) in utils_icl.py.

---

🚀 Inference

sh run_predict.sh

Bibinfo

If you found this repo helpful, please help us by citing this paper:

@inproceedings{li-etal-2025-decoding-graphs,
    title = "Decoding on Graphs: Faithful and Sound Reasoning on Knowledge Graphs through Generation of Well-Formed Chains",
    author = "Li, Kun  and Zhang, Tianhua  and Wu, Xixin  and Luo, Hongyin  and Glass, James R.  and Meng, Helen M.",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.acl-long.1186/",
    doi = "10.18653/v1/2025.acl-long.1186",
    pages = "24349--24364",
    ISBN = "979-8-89176-251-0",
}