Research map / Time series / Foundation models for time series
A Unified Shape-Aware Foundation Model for Time Series Classification
Zhen Liu, Yucheng Wang, Boyuan Li, Junhao Zheng, Emadeldeen Eldele, Min Wu, Qianli Ma
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 |
|---|---|---|---|
| pretrain on large multi-domain data before fine-tuning (ablation) | UniShape · 18 UCR | avg accuracy: 0.8365 -> 0.8529 | Improved |
| add shape-aware adapter (ablation) | UniShape · 18 UCR | avg accuracy: 0.8428 -> 0.8529 | Improved |
| use CNN instead of Transformer in adapter (ablation) | UniShape · 18 UCR | avg accuracy: 0.8446 -> 0.8529 | Improved |
| use CNN instead of MLP in adapter (ablation) | UniShape · 18 UCR | avg accuracy: 0.8431 -> 0.8529 | Improved |
| add instance-prototype contrastive loss (ablation) | UniShape · 18 UCR | avg accuracy: 0.8444 -> 0.8529 | Improved |
| add shape-prototype contrastive loss (ablation) | UniShape · 18 UCR | avg accuracy: 0.8470 -> 0.8529 | Improved |
| add both prototype losses (ablation) | UniShape · 18 UCR | avg accuracy: 0.8411 -> 0.8529 | Improved |
| use Transformer instead of Inception CNN as encoder (ablation) | UniShape · 18 UCR | avg accuracy: 0.8512 -> 0.8529 | No clear effect |
| use Transformer instead of MLP as encoder (ablation) | UniShape · 18 UCR | avg accuracy: 0.5651 -> 0.8529 | Improved |
| increase pretraining label ratio from 10% to 100% (ablation) | UniShape · 18 UCR | avg accuracy: 0.8529 -> 0.8588 | No clear effect |
| increase pretraining label ratio from 0% to 10% (ablation) | UniShape · 18 UCR | avg accuracy: 0.8395 -> 0.8529 | Improved |
| scale up pretraining data (UCR to ALL 1.89M) (ablation) | UniShape · 18 UCR | avg accuracy | 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.
- Mantis — Mantis: Lightweight Foundation Model for Time Series Classification (ICML 2026) · better on 2 of 2 shared results
- MOMENT — MOMENT: A Family of Open Time-series Foundation Models (ICML 2024) · better on 2 of 2 shared results
- TS2Vec — TS2Vec: Towards Universal Representation of Time Series (AAAI 2022) · 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
- Abstracted Shapes as Tokens - A Generalizable and Interpretable Model for Time-series Classification (NeurIPS 2024 · similarity 0.89)
- ShapeX: Shapelet-Driven Post Hoc Explanations for Time Series Classification Models (NeurIPS 2025 · similarity 0.88)
- Learning Soft Sparse Shapes for Efficient Time-Series Classification (ICML 2025 · similarity 0.88)
- Diffusion Language-Shapelets for Semi-supervised Time-Series Classification (AAAI 2024 · similarity 0.86)
- Large Pre-trained time series models for cross-domain Time series analysis tasks (NeurIPS 2024 · similarity 0.84)
- Shape analysis for time series (NeurIPS 2024 · similarity 0.83)
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.
