No, they don't force the model to use a json schema, they simply use the model to extract the data, and then they feed that OCR result into the pipeline further

No, they don't force the model to use a json schema, they simply use the model to extract the data, and then they feed that OCR result into the pipeline further to evaluate the OCR results against the ground truth, and this is where JSON schema is used, and also another model (gpt4o).