Explaining Time Series Forecasting with Horizon-Resolved Attribution

Recent advances in explaining time series (TS) models have produced methods that identify which past values a prediction depends on. However, most existing methods return a single importance vector, which assumes that every predicted step depends on the same past values. In this paper, we show that this assumption does not hold in practice, as different forecast steps consistently depend on different past values across a wide range of backbones and benchmarks. We further show that this dependence is low-dimensional, as the explanations of all steps are built from a few shared maps whose number does not grow with the forecast length. Motivated by this observation, we propose Horizon-Resolved eXplanation (HRX), which adds a horizon axis to the explanation, so that every forecast step receives its own importance map. HRX is a simple yet effective plug-in framework with three components: 1) a default estimator that reads these maps out of any differentiable forecaster without modifying the TS backbone, and which any existing attribution method can replace, 2) an evaluation protocol that validates the horizon axis by shuffling which map belongs to which forecast step, and 3) a rank criterion that predicts, before any deletion is run, whether the axis is worth resolving on a given series. Extensive experiments show the improvement comes from the horizon axis, not the estimator.