OCRBench provides the largest evaluation suite yet for OCR capabilities in large multimodal models, revealing gaps in multilingual, handwritten, and mathematical text handling.
Llavar: Enhanced visual instruction tuning for text-rich image understanding, 2023
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
2023 1verdicts
ACCEPT 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
OCRBench: On the Hidden Mystery of OCR in Large Multimodal Models
OCRBench provides the largest evaluation suite yet for OCR capabilities in large multimodal models, revealing gaps in multilingual, handwritten, and mathematical text handling.