Benchmark RadarAI BENCHMARK PROFILE
RealDocBench
RealDocBench evaluates field-level QA and layout understanding on real regulated documents, with 1,356 field-level questions over 581 documents and 1,500 annotated page images, scored on per-field accuracy and adjacency-aware layout metrics.
- Released
- 2026-06-05
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
It addresses the gap in document parsing evaluation by focusing on real-world regulated documents and specific field-level needs, enabling cost-aware comparisons of commercial and open-source systems.
Motivation
Document parsing systems are increasingly deployed in high-stakes, regulated workflows such as mortgage underwriting, financial reporting, supply-chain logistics, and clinical records.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.