Datalab Releases lift: A 9B Open-Weights Vision Model That Extracts Structured JSON From PDFs Using Schemas

Datalab has released lift, a 9-billion parameter open-weights model designed to convert documents and images into structured JSON files based on specific user-defined schemas. The model aims to improve data extraction accuracy by using schema-constrained decoding and a trained abstention mechanism that returns null values rather than hallucinating missing information.