anonymous-8421/VL-DocIR-RepresentativeSubset

Representative Subset Creation This dataset represents a representative subset of VL-DocIR dataset (https://huggingface.co/datasets/anonymous-8421/VL-DocIR). Subset creation method: Random sampling of 5 queries for the combination of each data source and evidence structure type (we ensure that no query is taken more than once). This results in 120 queries. Collection of all documents that are referenced by query evidence. Abstract VL-DocIR is a page-level… See the full description on the dataset page: https://huggingface.co/datasets/anonymous-8421/VL-DocIR-RepresentativeSubset.