Discover how Query, Key, and Value projections enable LLMs to process context. Learn the math behind attention mechanisms and what these matrices actually learn.