OmniDocBench
OmniDocBench is a comprehensive benchmark for document understanding tasks, designed to evaluate models' parsing and comprehension abilities across diverse document types such as PDFs, scanned documents, tables, and mixed text-image layouts. The benchmark covers multiple subtasks including OCR recognition, layout analysis, and information extraction, providing a standardized performance evaluation framework for document understanding systems and widely used in academic research and model capability comparison.
Timeline (last 90 days)
OXR officially claims to rank first simultaneously on three mainstream document parsing benchmarks: olmOCR-Bench, OmniDocBench v1.6, and DocParsingBench.
OmniDocBench is a comprehensive benchmark for full-scenario document parsing, covering multiple document types and element categories, and has been iterated into versions such as v1.6.
Youtu-Parsing-Omni has not yet entered the official OmniDocBench leaderboard; its 96.96 score comes from the Tencent model card, while the older Youtu-Parsing's 93.74 score is on the official leaderboard.
In GitHub issue #14, the OmniDocBench team reran Youtu-Parsing on v1.5 and got an overall score of 93.37, while Tencent's internal number is 93.22; the formula CDM differs by about 2 points, and the table TEDS differs by about 2 points in the opposite direction.
All Facts (3)
OXR officially claims to rank first simultaneously on three mainstream document parsing benchmarks: olmOCR-Bench, OmniDocBench v1.6, and DocParsingBench.
50%UnverifiedOmniDocBench is a comprehensive benchmark for full-scenario document parsing, covering multiple document types and element categories, and has been iterated into versions such as v1.6.
50%UnverifiedIn GitHub issue #14, the OmniDocBench team reran Youtu-Parsing on v1.5 and got an overall score of 93.37, while Tencent's internal number is 93.22; the formula CDM differs by about 2 points, and the table TEDS differs by about 2 points in the opposite direction.
50%