Forecasting has a rich tradition in ML, spanning key areas such as time-series analysis, online learning, data-driven decisions and quantitative finance. Recent advances in foundation models, however, raise a qualitatively new question: can general-purpose AI systems reliably anticipate future events across diverse real-world domains? Indeed, forecasting is often viewed as a hallmark of sophisticated intelligence that requires internalizing patterns in dynamic environments and reasoning about consequences in the noisy real world, and we are witnessing a growing research efforts on advancing and benchmarking forecasting capabilities of AI systems.

Motivated by its deep roots and emerging paradigms, we envision forecasting as an exciting research program that requires lens from foundation models, agentic design, benchmarking, probabilistic reasoning, information retrieval, regret minimization, world modeling, etc. As the AI community seeks the next frontier in AI capabilities, this workshop aims to bring together researchers across machine learning, statistics, economics, finance and others to explore forecasting both as a foundational technical challenge and as a core capability of general-purpose AI systems.