Interpretable Oracle Bone Script Decipherment through Radical and Pictographic Analysis with LVLMs

Kaixin Peng, Mengyang Zhao, Haiyang Yu, Teng Fu, Bin Li*

This repository is the implementation of the paper "Interpretable Oracle Bone Script Decipherment through Radical and Pictographic Analysis with LVLMs".

Our dataset (PD-OBS) is avaliable here.

Dataset

Statistic

Component	Count	Description
Characters	47,157	Characters collected from KangXi Dictionary
Regular Images	47,157	Modern character forms collected from Handian
Ancient Images	10,968	Ancient character forms collected from Open Ancient Chinese Characters Glyphs Database
OBS Categories (Including duplicates)	11,739	OBS collected from HUST-OBC and EVOBC

Directory

PD-OBS/
  dictionary/
    dictionary.json
    id_to_char.json
  img/
    regular/
    ancient/
    obs/

• dictionary/dictionary.json: Authoritative Dictionary Definitions, radical analysis and pictographic analysis.
• dictionary/id_to_char.json/: ID-Character mapping.
• img/regular/: Modern Chinese character writing style images.
• img/ancient/: Ancient Chinese character writing style images.
• img/obs/: Corresponding oracle bone script images.

Train & Inference & Checkpoints

To be released upon paper publication.

Cite

@misc{peng2025interpretableoraclebonescript,
      title={Interpretable Oracle Bone Script Decipherment through Radical and Pictographic Analysis with LVLMs}, 
      author={Kaixin Peng and Mengyang Zhao and Haiyang Yu and Teng Fu and Bin Li},
      year={2025},
      eprint={2508.10113},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2508.10113}, 
}