PictOBI-20k, a new 15k-question benchmark, shows top large multimodal models reach only 53.7% accuracy at matching oracle bone pictographs to object photos, with vision encoders often outperforming the full models.
Obi-bench: Can lmms aid in study of ancient script on oracle bones?
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
PictOBI-20k: Unveiling Large Multimodal Models in Visual Decipherment for Pictographic Oracle Bone Characters
PictOBI-20k, a new 15k-question benchmark, shows top large multimodal models reach only 53.7% accuracy at matching oracle bone pictographs to object photos, with vision encoders often outperforming the full models.