LLM Interpretability

A research area focused on understanding the internal mechanisms of AI language models, examining how models make decisions and how information flows through their layers. Common methods include gradient attribution, attention visualization, and probing classifiers, with the goal of making model behavior more transparent and auditable.