This is the official release of the training dataset for paper Taxonomy-Aware Representation Alignment for Hierarchical Visual Recognition with Large Multimodal Models. Code is available at https://github.com/PKU-ICST-MIPL/TARA_CVPR2026.

	Data Source




	
	Training

We randomly sample 1-shot data per category from iNaturalist2021 dataset.

Data Fields

image: input image(s) data type: dict

problem: input question data type: string

label: coarse-to-fine categories… See the full description on the dataset page: https://huggingface.co/datasets/StevenHH2000/iNat21-1shot-fewshots.