Tag
1 article
Datalab introduces lift, a 9B open-weight vision model that accurately extracts structured JSON from PDFs using schema-constrained decoding and abstention training, achieving 90.2% field accuracy.
