ExtractBench: A Benchmark for Schema-Guided Enterprise Document Extraction
The benchmark scores extraction accuracy, completeness, grounding, and cost on 370 enterprise documents.
ExtractBench covers 4,869 pages across 8 business domains and 67 document types. The authors say commercial VLMs handle short documents well but often truncate long record lists. Coding agents keep higher accuracy at much higher cost. LlamaExtract Agentic Plus ranks first on the reported metrics, with coding-agent-level accuracy at lower cost. ArXiv · AI/CL/LG's note
ExtractBench covers 4,869 pages across 8 business domains and 67 document types. The authors say commercial VLMs handle short documents well but often truncate long record lists. Coding agents keep higher accuracy at much higher cost. LlamaExtract Agentic Plus ranks first on the reported metrics, with coding-agent-level accuracy at lower cost. ArXiv · AI/CL/LG's note
score 5