Research map / Architectures and efficiency / Pruning and efficient tuning
Task-Adaptive Parameter-Efficient Fine-Tuning for Weather Foundation Models
Shilei Cao, Hehai Lin, Jiashun Cheng, Yang Liu, Guowen Li, Xuehe Wang, Juepeng Zheng, Haoyuan Liang, Meng Jin, Chengwei Qin, Hong Cheng, Haohuan Fu
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 |
|---|---|---|---|
| combine TADP and SFAS instead of either alone (ablation) | WeatherPEFT · Downscaling T2m | RMSE: 1.161 (SFAS only) -> 1.119 | Improved |
| add SFAS to TADP (versus TADP only) (ablation) | WeatherPEFT · Regional precipitation 12h | SEEPS: 0.549 -> 0.368 | Improved |
| add internal pattern extraction to TADP (ablation) | WeatherPEFT · Downscaling T2m (limited budget) | RMSE: 1.140 -> 1.119 | Improved |
| add external pattern integration to TADP (ablation) | WeatherPEFT · Downscaling T2m (limited budget) | RMSE: 1.130 -> 1.119 | Improved |
| add annealed randomness to Fisher-guided selection (ablation) | SFAS · Downscaling T2m (limited budget) | RMSE: 1.130 -> 1.119 | Improved |
| add internal pattern extraction (larger budget) (ablation) | WeatherPEFT · Downscaling T2m (increased budget) | RMSE: 0.970 -> 0.916 | Improved |
| increase selected parameter fraction k from 0.001 to 0.04 (ablation) | WeatherPEFT · Regional precipitation 12h | SEEPS: 0.368 -> 0.302 | Improved |
| increase LoRA rank (8 to 256) at fixed method (ablation) | LoRA · Regional precipitation 12h | SEEPS: 0.495 (r=8) -> 0.495 (r=256) | No clear effect |
| lengthen soft prompt from P=20 to P=40 or more (ablation) | TADP · Regional precipitation 12h | SEEPS: 0.368 -> 0.400 | Hurt |
| enlarge adapter hidden dims from (8,6,16) to (32,13,512) (ablation) | TADP · Regional precipitation 12h | SEEPS: 0.368 -> 0.376 | Hurt |
| vary initial decay factor gamma (1.0 to 0.2) (ablation) | SFAS · Regional precipitation 12h | SEEPS: 0.369 -> 0.368 | No clear effect |
| use WeatherPEFT instead of LoRA on a different backbone | Prithvi-WxC · Regional precipitation 12h | SEEPS: 0.647 -> 0.405 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DoRA — DoRA: Weight-Decomposed Low-Rank Adaptation (ICML 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- WeatherGFM: Learning a Weather Generalist Foundation Model via In-context Learning (ICLR 2025 · similarity 0.88)
- Parameter Efficient Fine-tuning via Explained Variance Adaptation (NeurIPS 2025 · similarity 0.85)
- WeatherSyn: An Instruction Tuning MLLM For Weather Forecasting Report Generation (ICML 2026 · similarity 0.85)
- Multi-Scale Finetuning for Encoder-based Time Series Foundation Models (NeurIPS 2025 · similarity 0.85)
- Be Confident in What You Know: Bayesian Parameter Efficient Fine-Tuning of Vision Foundation Models (NeurIPS 2024 · similarity 0.84)
- Real-World Adverse Weather Image Restoration via Dual-Level Reinforcement Learning with High-Quality Cold Start (NeurIPS 2025 · similarity 0.84)
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.
