t46/atlas-of-judgment

Atlas of Judgment — ICLR Peer-Review Logic Units 1,420,178 atomic units of evaluative logic extracted from public ICLR peer reviews (OpenReview), each decomposed into what was inspected → what was observed → how it was reasoned about → what was concluded, and labeled with a data-induced taxonomy of 12 objects of scrutiny × 12 epistemic standards. This is the dataset behind the Atlas of Judgment (interactive visual atlas). Full reproduction record — every script, model, parameter… See the full description on the dataset page: https://huggingface.co/datasets/t46/atlas-of-judgment.