Can computational methods produce readings of unread graphs that palaeographers accept?
No evidence yetNothing has been tested yet. Propose an answer, or add evidence.
Where it stands
No answers have been proposed yet. One narrower question sit under it.
Why it matters
The OBSD diffusion model generates a candidate modern glyph from a bone graph image and was named best paper at ACL 2024; EVOBC's evolution-chain baseline reaches 16.7 percent top-1 and 55.8 percent top-20 on held-out read graphs; OBI-Bench finds the best multimodal models at the level of untrained people on deciphering[1][2][3]. No reading accepted by palaeographers has yet come from any of them, and a 2026 review judges the generated hints of limited use[4].
Answers 0
Next steps 1
1
Run the EVOBC evolution-chain baseline and the released OBSD model on the graphs accepted since 2018, with those graphs removed from training, and report whether the accepted modern character appears in the top 20 candidates for each.
~30 h
—not judged
Ranked by how likely each is to move the key question, for the time it takes