Research map / Time series / Foundation models for time series
Channel Adapter for Time Series Foundation Models in Zero-Shot Multivariate Forecasting
Dongyuan Li, Renhe Jiang, Shun Zheng, Zheng Dong, Haotian Gao, Ying Zhang, Jiang Bian
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 ChaTSFM to Moirai | Moirai · Solar (zero-shot) | MSE: 0.515 -> 0.394 | Improved |
| add ChaTSFM to Chronos | Chronos · Solar (zero-shot) | MSE: 1.922 -> 1.496 | Improved |
| add ChaTSFM to Moirai | Moirai · ETTm1 (zero-shot) | MSE: 0.950 -> 0.808 | Improved |
| add ChaTSFM to Moirai 2.0 with input length 512-3072 | Moirai 2.0 · ETTm1 (zero-shot, long input) | MSE: 0.436 -> 0.368 | Improved |
| add pre-training to ChaTSFM (ablation) | ChaTSFM · Solar (full-shot) | MSE: 0.242 -> 0.228 | Improved |
| add curated data selection instead of random selection (ablation) | ChaTSFM pre-training · Solar (zero-shot) | MSE: 0.435 -> 0.394 | Improved |
| add curated selection instead of full dataset (ablation) | ChaTSFM pre-training · Solar (zero-shot) | MSE: 0.444 -> 0.394 | Improved |
| add learned domain-adaptive metric instead of cosine similarity (ablation) | ChaTSFM · Solar (zero-shot) | MSE: 0.512 -> 0.394 | Improved |
| add hyper-network metric instead of instance-wise MLP (ablation) | ChaTSFM · Solar (zero-shot) | MSE: 0.418 -> 0.394 | Improved |
| add Top-K sparse aggregation instead of full attention (ablation) | ChaTSFM · Solar (zero-shot) | MSE: 0.435 -> 0.394 | Improved |
| add channel adapter instead of channel independent (ablation) | ChaTSFM · Solar (zero-shot) | MSE: 0.515 -> 0.394 | Improved |
| add adaptive Top-K instead of fixed Top-5 (ablation) | ChaTSFM · Solar (zero-shot) | MSE: 0.402 -> 0.394 | Improved |
Closest papers by meaning
- Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Series (NeurIPS 2024 · similarity 0.90)
- ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Data (AAAI 2025 · similarity 0.89)
- Multi-Scale Finetuning for Encoder-based Time Series Foundation Models (NeurIPS 2025 · similarity 0.89)
- AdaPTS: Adapting Univariate Foundation Models to Probabilistic Multivariate Time Series Forecasting (ICML 2025 · similarity 0.89)
- From Similarity to Superiority: Channel Clustering for Time Series Forecasting (NeurIPS 2024 · similarity 0.88)
- TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecaster (NeurIPS 2025 · similarity 0.88)
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.
