Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Impermanent: A Live Benchmark for Temporal Generalization in Time Series Forecasting

Azul Garza, Renée Rosillo, Rodrigo Mendoza-Smith, David Salinas, Andrew Robert Williams, Arjun Ashok, Mononito Goswami, José Martín Juárez

TL;DR

Impermanent is introduced, a live benchmark that evaluates forecasting models under open-world temporal change by scoring forecasts sequentially over time on continuously updated data streams, enabling the study of temporal robustness, distributional shift, and performance stability rather than one-off accuracy on a frozen test set.

Abstract

Recent advances in time-series forecasting increasingly rely on pre-trained foundation-style models. While these models often claim broad generalization, existing evaluation protocols provide limited evidence. Indeed, most current benchmarks use static train-test splits that can easily lead to contamination as foundation models can inadvertently train on test data or perform model selection using test scores, which can inflate performance. We introduce Impermanent, a live benchmark that evaluates forecasting models under open-world temporal change by scoring forecasts sequentially over time on continuously updated data streams, enabling the study of temporal robustness, distributional shift, and performance stability rather than one-off accuracy on a frozen test set. Impermanent is instantiated on GitHub open-source activity, providing a naturally live and highly non-stationary dataset shaped by releases, shifting contributor behavior, platform/tooling changes, and external events. We focus on the top 400 repositories by star count and construct time series from issues opened, pull requests opened, push events, and new stargazers, evaluated over a rolling window with daily updates, alongside standardized protocols and leaderboards for reproducible, ongoing comparison. By shifting evaluation from static accuracy to sustained performance, Impermanent takes a concrete step toward assessing when and whether foundation-level generalization in time-series forecasting can be meaningfully claimed. Code and a live dashboard are available at https://github.com/TimeCopilot/impermanent and https://impermanent.timecopilot.dev.

Impermanent: A Live Benchmark for Temporal Generalization in Time Series Forecasting

TL;DR

Impermanent is introduced, a live benchmark that evaluates forecasting models under open-world temporal change by scoring forecasts sequentially over time on continuously updated data streams, enabling the study of temporal robustness, distributional shift, and performance stability rather than one-off accuracy on a frozen test set.

Abstract

Recent advances in time-series forecasting increasingly rely on pre-trained foundation-style models. While these models often claim broad generalization, existing evaluation protocols provide limited evidence. Indeed, most current benchmarks use static train-test splits that can easily lead to contamination as foundation models can inadvertently train on test data or perform model selection using test scores, which can inflate performance. We introduce Impermanent, a live benchmark that evaluates forecasting models under open-world temporal change by scoring forecasts sequentially over time on continuously updated data streams, enabling the study of temporal robustness, distributional shift, and performance stability rather than one-off accuracy on a frozen test set. Impermanent is instantiated on GitHub open-source activity, providing a naturally live and highly non-stationary dataset shaped by releases, shifting contributor behavior, platform/tooling changes, and external events. We focus on the top 400 repositories by star count and construct time series from issues opened, pull requests opened, push events, and new stargazers, evaluated over a rolling window with daily updates, alongside standardized protocols and leaderboards for reproducible, ongoing comparison. By shifting evaluation from static accuracy to sustained performance, Impermanent takes a concrete step toward assessing when and whether foundation-level generalization in time-series forecasting can be meaningfully claimed. Code and a live dashboard are available at https://github.com/TimeCopilot/impermanent and https://impermanent.timecopilot.dev.
Paper Structure (6 sections, 2 equations, 3 figures, 2 tables)

This paper contains 6 sections, 2 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Contributions.
  3. The Impermanent Benchmark
  4. Evaluation Protocol
  5. Results
  6. Conclusion and Future Work

Figures (3)

  • Figure 1: Weekly GitHub activity for a random sample of 25 repositories across four event types (issues opened, pull requests opened, push events, new stargazers). Top: weekly event counts. Bottom: cumulative counts over time.
  • Figure 2: Weekly-series summary statistics for a random sample of 25 repositories, grouped by event type (columns). Each panel shows a relationship between two descriptors: log low/high bandpower ratio (slow vs fast variation) vs permutation entropy (irregularity), and spectral centroid (typical frequency) vs spectral entropy (spectral spread).
  • Figure 3: Monthly event counts for a random sample of six repositories across four activity signals. The series exhibit hallmark features of real-world temporal data: intermittency (long zero-count stretches), burstiness (isolated spikes), level shifts, and heterogeneous scales across repositories and event types.