Research map / Time series / Foundation models for time series
Residual-Guided Multi-Resolution Refinement of Foundation Models: A Case Study in Drought Forecasting
Wentao Gao, Jiuyong Li, Lin Liu, Thuc Duy Le, Jixue Liu, Yanchang Zhao, Yun Chen
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 |
|---|---|---|---|
| add RGMR wrapper to frozen TimesFM | TimesFM v1.0 · SPEI Location 1 | MSE: 0.391 -> 0.318 | Improved |
| add RGMR wrapper to frozen TimeGPT-1 | TimeGPT-1 · SPEI Location 1 | MSE: 0.437 -> 0.325 | Improved |
| add RGMR wrapper to frozen TabPFN-ts | TabPFN-ts v1.0 · SPEI Location 1 | MSE: 0.462 -> 0.344 | Improved |
| use frozen foundation model instead of generic deep baselines | TimesFM v1.0 · SPEI Location 1 | MSE: 0.533 (DLinear) -> 0.391 | Improved |
| add coarse projection only (ablation) | frozen TimesFM v1.0 · SPEI avg 3 sites | MSE: 0.411 -> 0.387 | Improved |
| add multi-resolution proposals without refinement (ablation) | frozen TimesFM v1.0 · SPEI avg 3 sites | MSE: 0.387 -> 0.372 | Improved |
| add residual-guided updates (ablation) | TimesFM multi-resolution · SPEI avg 3 sites | MSE: 0.372 -> 0.341 | Improved |
| add adaptive residual weights (ablation) | RGMR (TimesFM) · SPEI avg 3 sites | MSE: 0.341 -> 0.334 | Improved |
| add adaptive residual weights (ablation) | RGMR (TabPFN-ts) · SPEI avg 3 sites | MSE: 0.380 -> 0.367 | Improved |
| RGMR refinement at finest stride r=1 (level 5) (ablation) | RGMR · South Australia test set | sample-wise improvement ratio: 48.5% | No clear effect |
| RGMR refinement from stride 12 to 6 (level 2) (ablation) | RGMR · South Australia test set | sample-wise improvement ratio: 39.7% | Hurt |
| change resolution scale from [12,6,3,2,1] to alternatives (ablation) | RGMR (TimesFM) · SPEI Location 1 | MSE: 0.318 -> 0.322 to 0.330 | No clear effect |
Closest papers by meaning
- Pushing the Limits of High-Resolution Weather Forecasting through Data Scaling (ECCV 2026 · similarity 0.85)
- Multi-Scale Finetuning for Encoder-based Time Series Foundation Models (NeurIPS 2025 · similarity 0.83)
- DroughtSet: Understanding Drought Through Spatial-Temporal Learning (AAAI 2025 · similarity 0.83)
- Quantile-Regression-Ensemble: A Deep Learning Algorithm for Downscaling Extreme Precipitation (AAAI 2024 · similarity 0.83)
- SEMPO: Lightweight Foundation Models for Time Series Forecasting (NeurIPS 2025 · similarity 0.83)
- Precipitation Downscaling with Spatiotemporal Video Diffusion (NeurIPS 2024 · similarity 0.83)
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.
