Research map / Time series / Foundation models for time series
TiMi: Empower Time Series Transformers with Multimodal Mixture of Experts
Jiafeng Lin, Yuxuan Wang, Huakun Luo, Jianmin Wang, Zhongyi Pei
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 MMoE plug-in to PatchTST | PatchTST · Time-MMD (6 datasets) | MSE: average promotion of 18.2% | Improved |
| add MMoE plug-in to TimeXer | TimeXer · Time-MMD SocialGood | MSE: 1.302 -> 0.962 | Improved |
| add MMoE plug-in to Autoformer (encoder-decoder, point-wise) | Autoformer · Time-MMD (6 datasets) | MSE: average promotion of 12.4% | Improved |
| add series-aware MoE (SMoE) on top of TMoE (ablation) | TiMi · Time-MMD SocialGood | MSE: 0.988 -> 0.886 | Improved |
| add SMoE on Traffic (ablation) | TiMi · Time-MMD Traffic | MSE: 0.154 -> 0.154 | No clear effect |
| add text-informed MoE (TMoE) on top of SMoE (ablation) | TiMi · Time-MMD Energy | MSE: 0.251 -> 0.229 | Improved |
| add TMoE on Traffic MAE (ablation) | TiMi · Time-MMD Traffic | MAE: 0.214 -> 0.218 | Hurt |
| use specialized TMoE+SMoE instead of vanilla MoE (ablation) | TiMi · Time-MMD Agriculture | MSE: 0.231 -> 0.193 | Improved |
| use TMoE instead of standard cross-attention for text (ablation) | TiMi · Time-MMD Energy | MSE: 0.272 -> 0.229 | Improved |
| use pre-trained LLM instead of random initialization (ablation) | TiMi · Time-MMD Health(US) | MSE: 1.293 -> 1.194 | Improved |
| use Qwen2.5-7B instead of smaller LLMs (Qwen2.5-0.5B, GPT-2) (ablation) | TiMi · Time-MMD Security | MSE: 73.2 / 73.1 -> 71.5 | Improved |
| add text via smaller LLMs vs unimodal PatchTST (ablation) | PatchTST · Time-MMD Security | MSE: 84.3 -> 73.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · better on 1 of 1 shared result
- TimesNet — TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis (ICLR 2023) · better on 1 of 1 shared result
- Autoformer — Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting (NeurIPS 2021) · 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
- Rethinking Post-Training Recipes for Multimodal Time-Series Forecasting (NeurIPS 2026 · similarity 0.91)
- Time-MMD: Multi-Domain Multimodal Dataset for Time Series Analysis (NeurIPS 2024 · similarity 0.91)
- GPT4MTS: Prompt-based Large Language Model for Multimodal Time-series Forecasting (AAAI 2024 · similarity 0.91)
- TimeXL: Explainable Multi-modal Time Series Prediction with LLM-in-the-Loop (NeurIPS 2025 · similarity 0.90)
- Empowering Time Series Analysis with Large-Scale Multimodal Pretraining (NeurIPS 2026 · similarity 0.90)
- TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment (AAAI 2025 · similarity 0.90)
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.
