Research map / Time series / Foundation models for time series
R-Tuning: Wavelet-Decomposed Replay and Semantic Alignment for Continual Adaptation of Pretrained Time-Series Models
Tianyi Yin, Jingwei Wang, Chenze Wang, Han Wang, Jiexuan Cai, Min Liu, Yunlong Ma, Kun Gao, Yuting Song, Weiming Shen
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 |
|---|---|---|---|
| apply R-Tuning instead of raw model | Apollo · New task MAE | MAE (x10): 1.9 -> 1.0 | Improved |
| apply R-Tuning | Apollo · Old tasks MSE | MSE (x10): 6.9 -> 6.5 | Improved |
| vanilla fine-tuning on new data only | Apollo · Old tasks MSE | MSE (x10): 6.9 -> 7.8 | Hurt |
| add Wavelet-guided Replay (ablation) | R-Tuning (AutoTimes) · Old tasks | MAE: 0.55 ± 0.12 -> 0.52 ± 0.06 | Improved |
| add Semantic Alignment (ablation) | R-Tuning (AutoTimes) · Old tasks | MSE: 0.69 ± 0.11 -> 0.68 ± 0.09 | Improved |
| add Wavelet-guided Replay (ablation) | R-Tuning (AutoTimes) · New task | MAE: 0.13 ± 0.01 -> 0.11 ± 0.04 | Improved |
| add Semantic Alignment (ablation) | R-Tuning (AutoTimes) · New task | MSE: 0.09 ± 0.03 -> 0.09 ± 0.04 | No clear effect |
| increase synthetic replay ratio from 1% to 5% (ablation) | R-Tuning · avg over six models | MAE: 0.1455 -> 0.1050 | Improved |
| increase synthetic replay ratio beyond 5% (ablation) | R-Tuning · avg over six models | MAE and MSE | No clear effect |
| add R-Tuning vs RanPAC replay baseline | Chronos · Old tasks MAE | MAE: 5.2±0.3/+1.51% vs 5.2±1.1/+2.83% | Mixed |
Closest papers by meaning
- Towards a General Time Series Forecasting Model with Unified Representation and Adaptive Transfer (ICML 2025 · similarity 0.88)
- Rethinking Post-Training Recipes for Multimodal Time-Series Forecasting (NeurIPS 2026 · similarity 0.87)
- Large Pre-trained time series models for cross-domain Time series analysis tasks (NeurIPS 2024 · similarity 0.86)
- Align-RAG: Alignment Is All You Need for TSFM In-Context Learning (NeurIPS 2026 · similarity 0.86)
- Learning the Context of Errors: Black-Box Online Adaptation of Time Series Foundation Models (NeurIPS 2026 · similarity 0.86)
- Less is More: Unlocking Specialization of Time Series Foundation Models via Structured Pruning (NeurIPS 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.
