TopoPrimer: The Missing Topological Context in Forecasting Models

TopoPrimer adds global topological context to time-series forecasting models using persistent homology and spectral sheaf coordinates.

The framework reports consistent accuracy gains across four benchmarks, including improvements for Chronos, TimesFM, and ECL-related settings.

It is especially robust in difficult regimes such as seasonal demand spikes and cold-start scenarios.

The result suggests topology-aware signals can strengthen forecasting beyond standard model inputs.