It's TIME: Towards the Next Generation of Time Series Forecasting Benchmarks

The 'It's TIME' project addresses a critical pain point in AI research: the 'circularity' of time series benchmarks where models are evaluated on the same legacy datasets (like ETT or Weather) that they were trained on, or which have been structurally leaked into training sets. While the technical contribution is necessary for the field to progress toward reliable Time Series Foundation Models (TSFMs), the project currently lacks adoption signals (0 stars, though 10 forks suggest some peer review activity). Its defensibility is low because its value is entirely dependent on community consensus; if major labs (Amazon, Google, Salesforce) do not adopt this specific methodology for their next-gen models (like Chronos or MOIRAI), the project remains a purely academic exercise. Frontier labs are a medium risk here because they frequently develop proprietary 'clean' datasets for internal validation that they may or may not open-source, often rendering external benchmarks secondary to their internal leaderboards.