Why your LLM fixates on weird words and how visualizing attention helps decode it. Explore how attention mechanisms reveal model focus and improve understa