Report
SoMark open-sources OXR-1.0, a 1.3B-parameter document parser
The announcement claims OXR tops olmOCR-Bench, OmniDocBench and DocParsingBench, and leads open-source models in parsing speed.
TLDR
SoMark open-sourced OXR-1.0, a 1.3B-parameter document parser that turns PDFs and page images into Markdown or JSON, according to a detailed account of the release. SoMark’s announcement claims first place on three benchmarks and the fastest parsing among open-source models. The account says a recursive self-improvement loop uses OXR’s parsing mistakes to generate training data without human labeling.
Combined views
131
5 Sources, first seen ago
