As the world’s leading conference on machine learning and artificial intelligence, NeurIPS provides one of the clearest windows into how frontier AI research is evolving - what fields are rising, which regions are driving progress, and where new scientific breakthroughs are emerging.
Our analysis of all accepted papers in 2024 and 2025 reveals the rise of Reinforcement Learning and Robotics, now the dominant and fastest-growing field at NeurIPS. This reflects a broader and renewed push toward physical AI, embodied intelligence, autonomous agents, and systems that can learn through interaction rather than static data. China has now overtaken all other regions as the largest contributor to NeurIPS - and continues to grow faster than any other geography, accelerating research output across a wide range of topics and reshaping the global balance of frontier AI. AI for Science is experiencing a major boom, driven by breakthroughs in protein folding, physics-informed neural networks (PINNs), climate modeling, and large-scale simulation tools that are bringing AI deeper into the scientific method itself.
Europe distinguishes itself with a clear specialization in Explainable AI, focusing on transparency, interpretability, and safety, an area where European labs now play an outsized global leadership role in relative terms. And importantly, beyond the US, China, and Europe, the rest of the world is rapidly narrowing the gap in excellent research, with strong research hubs in Singapore, Hong-Kong, South Korea, Canada, Australia and Japan increasingly present in frontier AI work and contributing to a more globally distributed AI research ecosystem.