TSFM
A model that transfers the large-model pretraining paradigm to numerical sequence forecasting. It is pretrained on massive heterogeneous sequences to learn cross-domain temporal patterns and enable zero-shot inference.
Timeline (last 90 days)
FairAMO is a controlled evaluation protocol whose core design is to freeze the outputs of candidate TSFMs so that accuracy changes can only come from the orchestration strategy.
Time series foundation models (TSFMs) refer to models pretrained on large-scale multi-domain time series data that can be transferred to downstream forecasting tasks via zero-shot or lightweight fine-tuning; representative works include Chronos, TimesFM, Moirai, and others.
Different TSFMs differ significantly in performance on the same task, and even their ranking can flip depending on the task type.
All Facts (3)
FairAMO is a controlled evaluation protocol whose core design is to freeze the outputs of candidate TSFMs so that accuracy changes can only come from the orchestration strategy.
50%UnverifiedTime series foundation models (TSFMs) refer to models pretrained on large-scale multi-domain time series data that can be transferred to downstream forecasting tasks via zero-shot or lightweight fine-tuning; representative works include Chronos, TimesFM, Moirai, and others.
50%UnverifiedDifferent TSFMs differ significantly in performance on the same task, and even their ranking can flip depending on the task type.
50%