Invited talk “Understanding Reasoning in LLMs and (Agentic) LLMs” at the ICML 2026 Workshop on Compositional Learning: Safety and Agents :sunny: