Frontiers in AI Reasoning and Applications

Research Project, 2026

Recent advances in AI reasoning, exemplified by large reasoning models such as GPT-5, merge the capabilities of large language models with mechanisms like chain-of-thought (CoT) and formal verification tools. These breakthroughs are reshaping problem-solving across domains, from mathematics to autonomous agents. Landmark achievements, such as AI systems winning gold medals at the International Mathematical Olympiad, demonstrate their potential to match or surpass top human performance in demanding intellectual tasks.This planning project brings together leading researchers in AI and formal methods to establish an excellence cluster devoted to advancing the foundations of AI reasoning and exploring its applications. During the planning phase, leveraging our international network of world-leading researchers in the area, we will build a research consortium spanning academia and industry and define the cluster’s scientific agenda. Key research themes include reinforcement learning for complex reasoning; neurosymbolic AI integrating intuitive LLM reasoning with formal symbolic methods; cognitive science–inspired architectures for abstraction and adaptability; interpretable AI for safety-critical domains; and resource-efficient reasoning. The cluster will also identify and investigate application areas, with AI for mathematics as a flagship focus.