Contamination MapTime series / Model
NuTime
- Organisation
- Peking University / Microsoft
- Family
- NuTime
- Released
- 2024-11-12 (Checkpoint release date in the GitHub README news ('2024-11-12: Checkpoint of the self-supervised pretrained NuTime is released'), https://github.com/chenguolin/NuTime; arXiv v1 was 2023-10-11)
- Checkpoints
- ckpt/checkpoint_bias9.pth (GitHub chenguolin/NuTime)
- Weights
- Open · MIT
- Links
- PaperCode
Training data disclosure: Full list
Level determined for 100% of benchmark cells.
- TestTest data seen0
- SeenSeen, split unclear0
- TrainTraining data seen2
- Related data seen0
- NoNot seen47
- ?Unknown0
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Other declared training data
-
UCR Time Series Classification Archive (2018) · split: train of ts-classification
All 128 UCR datasets, official TRAIN splits only.
Evidence
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “In total, there are 60,555 in the training set of these 128 sub-datasets by their official split.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
UEA Multivariate Time Series Classification Archive (2018) · split: train of ts-classification
All 30 UEA datasets, training portions split into univariate channels.
Evidence
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “Multivariate data in the UEA archive is partitioned into univariate sequences for joint self-supervised pretraining. This finally leads to 1,386,874 sequences for training.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
Epilepsy (Andrzejak et al., 2001) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
SleepEEG (Kemp et al., 2000) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
HAR (Anguita et al., 2013) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
Gesture (Liu et al., 2009) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
FD-A (Lessmeier et al., 2016) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
FD-B (Lessmeier et al., 2016) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
ECG (Clifford et al., 2017) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
EMG (Goldberger et al., 2000) · split: train
Evidence
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 “The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15
Declared exclusions
-
Dataset UCR Time Series Classification Archive (2018) · test portion excluded
Evidence
“The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
Dataset UEA Multivariate Time Series Classification Archive (2018) · test portion excluded
Evidence
“The original training and testing splits of these datasets are retained, and only the training portions are merged.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15
Benchmark cells
GIFT-Eval
Not seen Jena Weather (GIFT-Eval) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen BizITObs - Application · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen BizITObs - Service · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen BizITObs - L2C · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Bitbrains - Fast Storage · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Bitbrains - rnd · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Restaurant (Recruit Restaurant Visitor Forecasting) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen ETT1 (GIFT-Eval) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen ETT2 (GIFT-Eval) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Loop Seattle (LibCity) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen SZ-Taxi (LibCity) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen M_DENSE (LibCity) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Solar (LSTNet, 137 Alabama PV plants) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Hierarchical Sales (Mancuso et al., 2020) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen M4 Yearly · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen M4 Quarterly · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen M4 Monthly · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen M4 Weekly · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen M4 Daily · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen M4 Hourly · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Hospital · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen COVID Deaths · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen US Births · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Saugeen River Flow · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Temperature Rain · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen KDD Cup 2018 · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Car Parts · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen ElectricityLoadDiagrams20112014 (UCI) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Long-horizon forecasting (ETT, Electricity, Traffic, Weather, Exchange, ILI)
Not seen ETTh1 · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen ETTh2 · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen ETTm1 · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen ETTm2 · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Electricity (ECL, 321 clients) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Traffic (PeMS San Francisco Bay, 862 sensors) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Weather (Jena, 2020; Autoformer) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Exchange-Rate · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen ILI (CDC influenza-like illness, national) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Time-series anomaly detection
Not seen TSB-AD-U (TSB-AD univariate) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen TSB-AD-M (TSB-AD multivariate) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen UCR Time Series Anomaly Archive (Hexagon ML/UCR, KDD Cup 2021) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Server Machine Dataset (SMD) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Mars Science Laboratory (MSL) telemetry anomaly dataset · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Soil Moisture Active Passive (SMAP) telemetry anomaly dataset · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Pooled Server Metrics (PSM) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Secure Water Treatment (SWaT) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Yahoo S5 Labeled Anomaly Detection Dataset (Yahoo Webscope S5) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Not seen Numenta Anomaly Benchmark (NAB) · absent from the complete list of training data
Evidence (3)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“include: Epilepsy (Andrzejak et al., 2001), SleepEEG (Kemp et al., 2000), HAR (Anguita et al., 2013), Gesture (Liu et al., 2009), FD-A (Lessmeier et al., 2016), FD-B (Lessmeier et al., 2016), ECG (Clifford et al., 2017) and EMG (Goldberger et al., 2000).”
“The merged dataset consists of approximately 1.89 million univariate sequences for training.”
Time-series classification
Training data seen UCR Time Series Classification Archive (2018) · only the training split was used, not the benchmark's test data; the test portion was explicitly excluded
Evidence (4)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“The original training and testing splits of these datasets are retained, and only the training portions are merged.”
“In total, there are 60,555 in the training set of these 128 sub-datasets by their official split.”
“The original training and testing splits of these datasets are retained, and only the training portions are merged.”
Training data seen UEA Multivariate Time Series Classification Archive (2018) · only the training split was used, not the benchmark's test data; the test portion was explicitly excluded
Evidence (4)
“We consider three main sources: (1) UCR time series archive (Dau et al., 2019), (2) UEA time series archive (Bagnall et al., 2018) and (3) eight additional datasets used in recent technical papers (Eldele et al., 2021b; Zhang et al., 2022).”
“The original training and testing splits of these datasets are retained, and only the training portions are merged.”
“Multivariate data in the UEA archive is partitioned into univariate sequences for joint self-supervised pretraining. This finally leads to 1,386,874 sequences for training.”
“The original training and testing splits of these datasets are retained, and only the training portions are merged.”
Notes
BYOL self-supervised classification FM (~2M params). Evaluated by fine-tuning on UCR/UEA test sets; their training splits were in pre-training (also stated by the Mantis authors, arXiv 2502.15637v2 App. B.1). Repository later moved to microsoft/NuTime.
Sources
- NuTime (TMLR 2024, arXiv v3) · paper · retrieved 2026-10-15
- chenguolin/NuTime README · readme · retrieved 2026-10-15