Research map / Time series / Foundation models for time series
Adapt Data to Model: Adaptive Transformation Optimization for Domain-shared Time Series Foundation Models
Yunzhong Qiu, Zhiyao Cen, Zhongyi Pei, Chen Wang, Jianmin Wang
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 TATO transformation optimization to frozen LTMs | six frozen LTMs (Timer, Moirai, Chronos) · 8 datasets, average | MSE: 0.3901 -> 0.3372 (13.6% avg) | Improved |
| apply TATO to Timer-LOTSA | Timer-LOTSA · Exchange | MSE: 0.8317 -> 0.2881 | Improved |
| apply TATO to Timer-UTSD | Timer-UTSD · Weather | MSE: 0.4937 -> 0.6048 | Hurt |
| apply TATO to Moirai-large | Moirai-large · Exchange | MSE: 0.2450 -> 0.2635 | Hurt |
| apply TATO to Chronos-tiny | Chronos-tiny · ETTh2 | MSE: 0.4181 -> 0.4297 | Hurt |
| apply TATO on top of universally finetuned LTMs | Timer-UTSD/LOTSA finetuned · 8 datasets average | MSE: 0.3142 -> 0.2913 (7.3%) | Improved |
| increase number of optimization trials 50 -> 500 (ablation) | TATO · 3 LTMs, horizon 96 | MSE reduction (%): 7.2% -> 9.4% | Improved |
| increase number of samples 50 -> 500 (ablation) | TATO · 3 LTMs, horizon 96 | MSE reduction (%): 4.6% -> 8.8% | Improved |
| include Trimmer and Scaler operators (ablation) | TATO · 3 LTMs, horizon 96 | MSE reduction: no number given | Improved |
| include Denoiser operator and TwoStageRank step (ablation) | TATO · 3 LTMs, horizon 96 | mean vs median MSE reduction: removal gives higher mean but lower median, higher variance | Mixed |
Closest papers by meaning
- SEMPO: Lightweight Foundation Models for Time Series Forecasting (NeurIPS 2025 · similarity 0.88)
- AdaPTS: Adapting Univariate Foundation Models to Probabilistic Multivariate Time Series Forecasting (ICML 2025 · similarity 0.88)
- Towards a General Time Series Forecasting Model with Unified Representation and Adaptive Transfer (ICML 2025 · similarity 0.88)
- Less is More: Unlocking Specialization of Time Series Foundation Models via Structured Pruning (NeurIPS 2025 · similarity 0.87)
- ARM: Refining Multivariate Forecasting with Adaptive Temporal-Contextual Learning (ICLR 2024 · similarity 0.87)
- TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecaster (NeurIPS 2025 · similarity 0.87)
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.
