danielhjerresen/

BDS_M4_exam_gold_dataset

Dataset of 100 patent claims labeled using a Multi-Agent System. Pipeline:

Advocate agent (argues green) Skeptic agent (detects greenwashing) Judge agent (final decision)

Human review applied only when agents disagreed. Labels: is_green_gold = 1 (green) is_green_gold = 0 (not green)