Research on how architectural components like activation functions affect safety and security properties of LLMs

This presentation covers 5 research papers on large language models applied to Security of LLM Activation Functions and Architecture.