This dataset is llava style multiturn vlm dataset from jp1924/KoreanImageCaptioningDataset.
ko-vlm/jp1924_multiturn_kvqa · main · files are served by the source, never re-hosted here
