Roam ResearchfromHackernoon1 year agoEmpirical Results: GPT-2 Analysis of Transformer Memorization & Loss | HackerNoonValidation of the radius hypothesis in GPT-2 experiments enhances understanding of next-token prediction accuracy.