English translation is being prepared.

Understand this story

Story summary

The English summary is being updated. See the reporting timeline below.

Latest development · ReportOct 23 00:11

1 confirmed developments

Browse reports and discussion by time. Their evidence levels remain distinct.

00:11
MarkTechPost
OmniParseBench Developer Guide: Leaderboard Benchmarks and OCR Testing
Datalab's open-source document parsing benchmark OmniParseBench covers 2,937 pages of documents with 16,288 pass/fail tests and scores 27 parsers. Hands-on testing found that the top two parsers are statistically tied, the top spot changes hands across most document-type slices, and one published figure does not match the public files. The author also ran the tests themselves on a 50-page sample using Tesseract and Indra OCR.