MR-MedSeg is a large-scale dataset for Multi-Round Entity-Level Medical
Reasoning Segmentation (MEMR-Seg). It is designed for medical segmentation
dialogues in which later user queries may refer to masks produced in earlier
rounds, requiring cross-round entity-level reasoning rather than isolated
single-turn segmentation.
The dataset follows the construction idea described in the MediRound paper:
medical entities are selected from public medical segmentation resources… See the full description on the dataset page:
https://huggingface.co/datasets/Carryyy/MR-MedSeg.
