This dataset accompanies the paper "Identifying the Measurement Gap in
Remote Sensing VQA with a GSD-Sensitive Taxonomy" (IEEE GRSL, under review).
It provides per-task D / M1 / M2 taxonomy labels, inter-annotator agreement
(IAA) data, and reproducibility artifacts for four public RS-VQA benchmarks.
The taxonomy partitions tasks via a single counterfactual: if the GSD were
doubled, would the answer value change (M1)… See the full description on the dataset page:
https://huggingface.co/datasets/ganghyunnnn/rs-taxonomy-labels.
