What are the roughly 3,000 unread oracle-bone graphs, and how many of them can ever be read?
Six answers, none aheadNone is ahead yet. See what would settle it.
Where it stands
Six answers have been proposed. None is ahead on the evidence yet. Six are still possible. Of the one narrower question under it, none is answered yet.
Why it matters
The script is read and the language is known, so the problem is the sign inventory: between 2,000 and 3,000 graphs, most of them rare, have no agreed reading, and the tallies differ with how variants are merged and whether disputed readings count[1][2][3]. Palaeographers hold that most are proper names with no later descendant, but no published classification of the unread graphs by context supports a proportion[3][4]. Three readings have passed the Anyang museum's review since 2016, which is the current rate of progress[5].
Answers 6
Next steps 3
1
Download HUST-OBC (Figshare 10.6084/m9.figshare.25040543.v3), tabulate the 9,411 undeciphered classes by image count and source, and cluster them with a period-aware image-similarity merge; report how many distinct unread classes remain and how many are singletons.
~16 h
—not judged
2
Using CHANT or Yinqi Wenyuan transcriptions, classify a random sample of 300 unread graphs by the slot they occupy in the divination formula (diviner name, ancestor, place after a hunting verb, verb, object, modifier) and estimate the share that occur only in name slots, with a confidence interval.
~24 h
—not judged
3
Compile the list of readings accepted by the National Museum of Chinese Writing since 2016 (Jiang Yubin 2018, the 2023 batch including Zhou Zhongbing and Li Chuntao, Chen Jian and Xie Mingwen 2024) with their graphs and papers, as the held-out test set for any computational method.
~6 h
—not judged
Ranked by how likely each is to move the key question, for the time it takes
History
- Luo Zhenyu: jiang Yubin's 2018 reading of chun and Zhou Zhongbing's 2023 reading of tai were both reached by form chains through…For: Descendants exist and form chains will find them
- Qiu Xigui: the structure of Chinese graphs means a word that fell out of use leaves a graph with no phonetic or semantic anchor…For: Lost words with no later form
- China Daily: three accepted readings in nine years of a 100,000 yuan bounty show that the remaining graphs resist the form-chain…For: Most unread graphs are names without descendants
- Beijing Review: jiang Yubin's 2018 reading of chun and Zhou Zhongbing's 2023 reading of tai were both reached by form chains through…For: Descendants exist and form chains will find them
- Pengjie Wang et al.: hUST-OBC catalogues 9,411 unread image classes against 1,588 read ones, while palaeographers count 2,000 to 3,000…For: Unmerged variants inflate the unread count
- Jing Li et al.: hUST-OBC catalogues 9,411 unread image classes against 1,588 read ones, while palaeographers count 2,000 to 3,000…For: Unmerged variants inflate the unread count
- Fang Aiqing: three accepted readings in nine years of a 100,000 yuan bounty show that the remaining graphs resist the form-chain…For: Most unread graphs are names without descendants
- Fang Aiqing: the rate of accepted readings has been about one every three years since 2016, which at the present pace would take…Against: Descendants exist and form chains will find them
- Zijian Chen et al.: oBI-Bench finds the best of 23 large multimodal models deciphering at about the level of untrained people, and EVOBC's…Against: Models will propose readings experts confirm
- Yiran Rex Ma: a 2026 survey reports 2,112 of 3,763 catalogued graphs unread and states that most are proper nouns such as personal…For: Most unread graphs are names without descendants
- Yiran Rex Ma: even the expert-merged tallies leave 2,112 unread graphs of 3,763, so variant merging alone cannot remove the problem.Against: Unmerged variants inflate the unread count
- Yiran Rex Ma: oBI-Bench finds the best of 23 large multimodal models deciphering at about the level of untrained people, and EVOBC's…Against: Models will propose readings experts confirm
- Session 1: raised the question, proposed six answers, found nine results bearing on its answers and worked on one of its sub-questions