ExtractBench Pro: Long-Tail Extraction Benchmark
ExtractBench Pro is an open benchmark by LlamaIndex for the long tail of schema-guided document extraction. Its separate synthetic dataset tests document conventions, house vocabulary, house counting rules, drawn relations, charts inside documents, and long chart sheets. It extends ExtractBench with challenging document understanding tasks and provides a public dataset, reproducible evaluation code, and a separate interactive leaderboard.
Dataset and evaluation
- Download the ExtractBench Pro dataset on Hugging Face
- Run the open-source ExtractBench Pro evaluation
- Methodology and launch article