MATH/IEDA Prof. CHEN Kani highlighted that the core challenge in AI agent research is achieving ability evolution through reinforcement learning.