Note ParseBench evaluates document parsing across tables, charts, content faithfulness, semantic formatting, and visual grounding. Live leaderboard: https://www.parsebench.ai/ Evaluation code and quick start: https://github.com/run-llama/ParseBench

Note ExtractBench evaluates schema-guided extraction from enterprise documents, including value accuracy and source grounding. Live leaderboard: https://www.extractbench.ai/ Evaluation code and quick start: https://github.com/run-llama/ExtractBench

Note ExtractBench Pro tests the long tail of document extraction with a separate dataset and leaderboard. Challenges include document conventions, house vocabulary, counting, drawn relations, and charts. Live leaderboard: https://www.extractbench.ai/pro Evaluation code: https://github.com/run-llama/ExtractBench#extractbench-pro

Note ParseBench: A Document Parsing Benchmark for AI Agents. Read the methodology behind the parsing benchmark; use the dataset and evaluation code above to reproduce evaluations. Live results: https://www.parsebench.ai/

Note ExtractBench: A Benchmark for Schema-Guided Enterprise Document Extraction. Read the methodology for evaluating extracted values and source grounding. ExtractBench Pro is available as a separate dataset above. Live results: https://www.extractbench.ai/