Research map / Time series / Causal and generative time series
Beyond Accuracy: Are Time Series Foundation Models Well-Calibrated?
Coen Adler, Yuxin Chang, Samar Abdi, Felix Draxler, Padhraic Smyth
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use time series foundation models instead of N-BEATS and ARIMA baselines | TSFMs · Six univariate datasets, 64-step zero-shot | PCE: Reviews 0.061/0.176 -> 0.012-0.015 | Improved |
| use TSFMs on Patents dataset | TSFMs · Patents | PCE: Reviews 0.012 -> Patents 0.154-0.259 | Hurt |
| replace quantile head with Gaussian prediction head (ablation) | TSFM backbones · Six datasets | PCE and CCE | Hurt |
| replace quantile head with Student's t or mixture head (ablation) | TSFM backbones · Six datasets | PCE | No clear effect |
| use trajectory AR method instead of branching AR (ablation) | Chronos-Bolt/TimesFM/Moirai 2.0 · Long-term forecasting L=256 | PCE | Improved |
| increase AR forecast horizon from 16 to 64 or 128 (ablation) | Autoregressive TSFMs · Long-term forecasting L=256 | PCE | Improved |
| use non-AR long-horizon models (TiRex, YingLong) instead of AR models | TiRex, YingLong · Long-term forecasting L=256 | PCE | Improved |
| extend prediction length from short to 64 steps ahead | TSFMs · Reviews, M5, Crime | PCE: close to or below 0.05 | No clear effect |
Closest papers by meaning
- MOMENT: A Family of Open Time-series Foundation Models (ICML 2024 · similarity 0.88)
- In-Context Fine-Tuning for Time-Series Foundation Models (ICML 2025 · similarity 0.87)
- OATS: Online Data Augmentation for Time Series Foundation Models (NeurIPS 2026 · similarity 0.86)
- Multi-Scale Finetuning for Encoder-based Time Series Foundation Models (NeurIPS 2025 · similarity 0.85)
- Parsimony or Capability? Decomposition Delivers Both in Long-term Time Series Forecasting (NeurIPS 2024 · similarity 0.85)
- SynTSBench: Rethinking Temporal Pattern Learning in Deep Learning Models for Time Series (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
