BoundingDocs: a Unified Dataset for Document Question Answering with Spatial Annotations
This provides a standardized resource for training and evaluating large language models on document AI tasks, though it is incremental as it builds on existing datasets.
The authors created a unified dataset for document question answering by combining public datasets and reformulating tasks like information extraction into QA format, and they explored prompting techniques with bounding box information to improve document comprehension in open-weight models.
We present a unified dataset for document Question-Answering (QA), which is obtained combining several public datasets related to Document AI and visually rich document understanding (VRDU). Our main contribution is twofold: on the one hand we reformulate existing Document AI tasks, such as Information Extraction (IE), into a Question-Answering task, making it a suitable resource for training and evaluating Large Language Models; on the other hand, we release the OCR of all the documents and include the exact position of the answer to be found in the document image as a bounding box. Using this dataset, we explore the impact of different prompting techniques (that might include bounding box information) on the performance of open-weight models, identifying the most effective approaches for document comprehension.