Causal Time Series
Summary
Causal structure appears as both a data-generation assumption and a reasoning task in the time-series cluster.
What The Wiki Currently Believes
- CauKer combines Gaussian-process kernel composition with structural causal models to generate synthetic, causally coherent time series.
- TimeOmni-1 includes causality discovery as one of the perception capabilities in TSR-Suite.
Evidence
CauKer uses causality to create pretraining data; TimeOmni-1 uses causality as a reasoning/evaluation target. Together they suggest causal structure is not optional if the goal is temporal understanding rather than curve fitting.
Relation To Foundation TSFM Agenda
This page maps directly to the causal/control slot in the Foundation Time-Series Model Research Agenda. Current local evidence is strongest for causal structure as synthetic-data prior or reasoning target; it is still weak for counterfactual action-conditioned rollout.
| Agenda slot | Verdict | Evidence | Missing pieces |
|---|---|---|---|
| Causal structure | partially closes | CauKer supplies causally structured synthetic generation; TimeOmni-1 includes causal discovery as a reasoning task. | Needs transfer evidence on real temporal systems and richer causal benchmarks. |
| Control and counterfactuals | insufficient evidence | The page does not yet establish candidate-action rollout or intervention consequence prediction. | Needs explicit actions, control inputs, interventions, and counterfactual evaluation. |
Open Questions
- How much causal correctness is needed for synthetic pretraining to transfer?
- Can models learn causality from synthetic templates without overfitting to template artifacts?