Research map / Time series / Foundation models for time series
Context parroting: A simple but tough-to-beat baseline for foundation models in scientific machine learning
Yuanzhao Zhang, William Gilpin
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 context parroting instead of foundation models | zero-shot forecasting of chaotic systems · dysts (135 systems, L=512) | sMAPE and attractor KL | Improved |
| Increase context length beyond 512 for Chronos (ablation) | Chronos · dysts | forecast error | No clear effect |
| Increase context length for context parroting (ablation) | context parroting · dysts | forecast error | Improved |
| Use short context (below 512) | Chronos vs parroting · dysts | forecast error | Mixed |
| Vary embedding dimension D of parroting (ablation) | context parroting · dysts | valid prediction time | No clear effect |
| Add noise up to 0.1 to trajectories (ablation) | context parroting vs foundation models · dysts | valid prediction time: 2.17 -> 1.04 | Hurt |
| Change sampling granularity (10, 30, 50 points per Lyapunov time) (ablation) | context parroting · dysts | valid prediction time: 4.70 / 2.15 / 1.41 | No clear effect |
| Use long forecast horizon with parroting instead of Chronos | context parroting · dysts long horizon | attractor KL: 0.679 -> 0.412 | Improved |
| Use parroting on real-world and high-dimensional SciML tasks | context parroting · ECG, Circuit, Kuramoto, Turbulence | MAE, KL | Mixed |
| Lengthen context for parroting on power spectrum (ablation) | context parroting · dysts | Hellinger distance of power spectrum | Improved |
Closest papers by meaning
- Zero-shot forecasting of chaotic systems (ICLR 2025 · similarity 0.91)
- In-Context Fine-Tuning for Time-Series Foundation Models (ICML 2025 · similarity 0.87)
- Less is More: Unlocking Specialization of Time Series Foundation Models via Structured Pruning (NeurIPS 2025 · similarity 0.87)
- Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization (ICML 2025 · similarity 0.86)
- Scaling Law for Time Series Forecasting (NeurIPS 2024 · similarity 0.86)
- Exploring Representations and Interventions in Time Series Foundation Models (ICML 2025 · similarity 0.86)
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.
