Absolute Zero is a groundbreaking AI paradigm that enables large language models to learn complex reasoning skills—such as coding and mathematics—entirely without human-curated data. Developed by researchers from institutions including Tsinghua University, UC Berkeley, and Google DeepMind, this approach represents a significant shift in AI training methodologies .