There is No "apple" in Timeseries: Rethinking TSFM through the Lens of Invariance
This addresses the performance gap in timeseries AI for researchers and practitioners, offering a foundational critique rather than an incremental improvement.
The paper argues that timeseries foundation models (TSFMs) underperform because they rely on web-scale data aggregation, which fails to capture core concepts like 'apple' in timeseries, unlike in NLP or CV. It proposes shifting to principled dataset design based on invariance coverage to improve generalization and emergent behavior.
Timeseries foundation models (TSFMs) have multiplied, yet lightweight supervised baselines and even classical models often match them. We argue this gap stems from the naive importation of NLP or CV pipelines. In language and vision, large web-scale corpora densely capture human concepts i.e. there are countless images and text of apples. In contrast, timeseries data is built to complement the image and text modalities. There are no timeseries dataset that contains the concept apple. As a result, the scrape-everything-online paradigm fails for TS. We posit that progress demands a shift from opportunistic aggregation to principled design: constructing datasets that systematically span the space of invariance that preserve temporal semantics. To this end, we suggest that the ontology of timeseries invariances should be built based on first principles. Only by ensuring representational completeness through invariance coverage can TSFMs achieve the aligned structure necessary for generalisation, reasoning, and truly emergent behaviour.