Contamination MapTime series foundation models
Which benchmark datasets could each model have seen during training? Every coloured square links to a verbatim quote from the model's paper, model card or repository. Grey means not enough has been disclosed to tell.
Contamination map
- TestTest data seen15
- SeenSeen, split unclear254
- TrainTraining data seen150
- Related data seen76
- NoNot seen1052
- ?Unknown658
Rows are models, grouped by family; columns are benchmark datasets. Hover a square for the reason, click it for the evidence. Empty outlines mark benchmarks a model cannot take (e.g. image benchmarks for text-only models).
Benchmarks
-
GIFT-Eval
28 datasets
-
Long-horizon forecasting (ETT, Electricity, Traffic, Weather, Exchange, ILI)
9 datasets
-
Time-series anomaly detection
10 datasets
-
Time-series classification
2 datasets
Models
| Model | Organisation | Released | Disclosure | Determined |
|---|---|---|---|---|
| Chronos (T5) | Amazon | 2024-03-13 | Full list | 100% |
| Chronos-2 | Amazon | 2025-10-20 | Full list | 100% |
| Chronos-2-Synth | Amazon | 2025-11-24 | Full list | 100% |
| Chronos-Bolt | Amazon | 2024-11-26 | Partial list | 18% |
| DADA (General Time Series Anomaly Detector with Adaptive Bottlenecks and Dual Adversarial Decoders) | East China Normal University / Aalborg University / Huawei Noah's Ark Lab | 2024-05-24 | Full list | 24% |
| TSPulse r1 | IBM Research | 2025-06-03 | Full list | 100% |
| Lag-Llama | Lag-Llama authors (ServiceNow Research, Mila, Morgan Stanley et al.) | 2023-10-12 | Full list | 100% |
| Mantis (Mantis-8M) | Huawei Noah's Ark Lab (Paris) | 2025-02-21 | Partial list | 4% |
| Mantis+ (MantisPlus) | Huawei Noah's Ark Lab (Paris) | 2026-02-19 | Full list | 100% |
| MantisV2 | Huawei Noah's Ark Lab (Paris) | 2026-02-19 | Full list | 100% |
| Moirai-1.0-R | Salesforce AI Research | 2024-02-04 | Full list | 100% |
| Moirai-1.1-R | Salesforce AI Research | 2024-06-14 | None | 0% |
| Moirai-2.0-R | Salesforce AI Research | 2025-08-06 | Partial list | 73% |
| Moirai-MoE-1.0-R | Salesforce AI Research | 2024-10-14 | Full list | 100% |
| MOMENT-1 | Auton Lab, Carnegie Mellon University | 2024-02-06 | Full list | 75% |
| NuTime | Peking University / Microsoft | 2024-11-12 | Full list | 100% |
| Time-MoE | Xiaohongshu Inc., Princeton University, Squirrel Ai Learning, Griffith University | 2024-09-24 | Full list | 57% |
| Time-RCD (TimeRCD) | Tsinghua University (thu-sail-lab) | 2025-10-18 | Full list | 100% |
| Sundial (Timer 3.0) | Tsinghua University (THUML) | 2025-02-02 | Partial list | 73% |
| Timer | Tsinghua University (THUML) | 2024-02-04 | Full list | 100% |
| Timer-S1 | Tsinghua University (THUML) & ByteDance | 2026-03-05 | Partial list | 65% |
| Timer-XL (timer-base-84m) | Tsinghua University (THUML) | 2024-10-07 | Partial list | 57% |
| TimeRadar | Singapore Management University (mala-lab) | 2026-02-22 | Full list | 18% |
| TimesFM 1.0 | Google Research | 2024-05-03 | Full list | 26% |
| TimesFM 2.0 | Google Research | 2024-12-30 | Full list | 44% |
| TimesFM 2.5 | Google Research | 2025-09-15 | Partial list | 4% |
| TimesFM 3.0 | Google Research | 2026-08-31 | Partial list | 4% |
| Granite-TimeSeries-TTM-R1 | IBM Research | 2024-04-05 | Full list | 100% |
| Granite-TimeSeries-TTM-R2 | IBM Research | 2024-10-08 | Full list | 100% |
| Granite-TimeSeries-TTM-R2.1 | IBM Research | 2025-01-31 | Full list | 100% |
| Granite-TTM-R3 | IBM Research | 2026-05-21 | Partial list | 61% |
| Tiny Time Mixer (TTM) Research-Use R2 | IBM Research | 2024-10-03 | Full list | 100% |
| TTM-R3 | IBM Research | 2026-02-09 | Partial list | 73% |
| TiRex | NXAI | 2025-05-29 | Full list | 100% |
| TiRex 1.1 (gifteval) | NXAI | 2025-10-06 | Partial list | 61% |
| TiRex-2 | NXAI | 2026-07-01 | Partial list | 0% |
| TiRex-2-fevbench (TiRex-2-f) | NXAI | 2026-07-01 | Full list | 100% |
| TiRex-2-gifteval-pretrain (TiRex-2-gp) | NXAI | 2026-07-01 | Partial list | 61% |
| TiRex-2-gifteval-zs (TiRex-2-g) | NXAI | 2026-07-01 | Full list | 100% |
| Toto (2024 technical report) | Datadog | 2024-07-10 | Partial list | 40% |
| Toto 2.0 | Datadog | 2026-05-19 | Vague | 75% |
| Toto-2.0-2.5B-FT | Datadog | 2026-05-19 | Partial list | 73% |
| Toto-Open-Base-1.0 | Datadog | 2025-05-20 | Partial list | 63% |
| UniTS | Harvard University (MIMS lab) | 2024-02-29 | Full list | 100% |
| UniTS few-shot anomaly detection checkpoints | Harvard University (MIMS lab) | 2024-02-29 | Full list | 100% |
Datasets
- Bitbrains - Fast Storage
- Bitbrains - rnd
- BizITObs - Application
- BizITObs - L2C
- BizITObs - Service
- Car Parts
- CDC FluView ILINet (LOTSA)
- COVID Deaths
- Electricity (ECL, 321 clients)
- Electricity Weekly (Monash)
- ElectricityLoadDiagrams20112014 (UCI)
- ETT1 (GIFT-Eval)
- ETT2 (GIFT-Eval)
- ETTh1
- ETTh2
- ETTm1
- ETTm2
- Exchange-Rate
- Hierarchical Sales (Mancuso et al., 2020)
- Hospital
- ILI (CDC influenza-like illness, national)
- Jena Weather (GIFT-Eval)
- KDD Cup 2018
- LargeST (PeMS California)
- Loop Seattle (LibCity)
- M4 Daily
- M4 Hourly
- M4 Monthly
- M4 Quarterly
- M4 Weekly
- M4 Yearly
- M_DENSE (LibCity)
- Mars Science Laboratory (MSL) telemetry anomaly dataset
- NREL Solar Power Data for Integration Studies (2006)
- Numenta Anomaly Benchmark (NAB)
- Pooled Server Metrics (PSM)
- Restaurant (Recruit Restaurant Visitor Forecasting)
- Saugeen River Flow
- Secure Water Treatment (SWaT)
- Server Machine Dataset (SMD)
- Soil Moisture Active Passive (SMAP) telemetry anomaly dataset
- Solar (LSTNet, 137 Alabama PV plants)
- SZ-Taxi (LibCity)
- Temperature Rain
- Traffic (PeMS San Francisco Bay, 862 sensors)
- Traffic Weekly (Monash)
- TSB-AD-M (TSB-AD multivariate)
- TSB-AD-U (TSB-AD univariate)
- UCR Time Series Anomaly Archive (Hexagon ML/UCR, KDD Cup 2021)
- UCR Time Series Classification Archive (2018)
- UEA Multivariate Time Series Classification Archive (2018)
- US Births
- Weather (Jena, 2020; Autoformer)
- Yahoo S5 Labeled Anomaly Detection Dataset (Yahoo Webscope S5)