A key-information-extraction set aimed squarely at the core job - given a document, return the specific values that matter.
We use it for: benchmarking field-level precision and recall - measuring how often we hallucinate a plausible-but-absent value.
Attribution
This is an unmodified fork of
nanonets/key_information_extraction, created by the Qwen team. All weights, files and behaviour are identical to upstream — we rehost it so our experiments stay reproducible and version-pinned. The original license and all credit remain with the Qwen team. If you want the canonical dataset, please use the original.