Contamination MapTime series / Model
Chronos-2
- Organisation
- Amazon
- Family
- Chronos · version 2
- Released
- 2025-10-20 (Repository README news entry '20 Oct 2025: Chronos-2 released' (https://raw.githubusercontent.com/amazon-science/chronos-forecasting/main/README.md) and GitHub release v2.0.0 published 2025-10-20; technical report arXiv v1 submitted 17 Oct 2025 (https://arxiv.org/abs/2510.15821))
- Checkpoints
- amazon/chronos-2, autogluon/chronos-2, autogluon/chronos-2-small
- Weights
- Open · apache-2.0
- Links
- PaperModel cardCode
Training data disclosure: Full list
Level determined for 100% of benchmark cells.
- TestTest data seen0
- SeenSeen, split unclear0
- TrainTraining data seen7
- Related data seen3
- NoNot seen39
- ?Unknown0
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Other declared training data
-
ElectricityLoadDiagrams20112014 (UCI) · split: unknown
Table 6 frequencies: 15min, 1H, 1W, 1D; source cited: Godahewa et al. (2021). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions). Table 6 lists 15min/1H/1W/1D versions; matched to the UCI source, of which the long-horizon Electricity is a derived version.
-
KDD Cup 2018 · split: unknown
Table 6 frequencies: 1H, 1D; source cited: Godahewa et al. (2021). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
M4 Daily · split: unknown
Table 6 frequencies: 1D; source cited: Makridakis et al. (2020). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
M4 Hourly · split: unknown
Table 6 frequencies: 1H; source cited: Makridakis et al. (2020). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
M4 Monthly · split: unknown
Table 6 frequencies: 1M; source cited: Makridakis et al. (2020). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
M4 Weekly · split: unknown
Table 6 frequencies: 1W; source cited: Makridakis et al. (2020). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Mexico City Bikes · split: unknown
Table 6 frequencies: 1H, 1D, 1W; source cited: Ansari et al. (2024). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Pedestrian Counts · split: unknown
Table 6 frequencies: 1H, 1D, 1W; source cited: Godahewa et al. (2021). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
NREL Solar Power Data for Integration Studies (2006) · split: unknown
Table 6 frequencies: 5min, 10min, 1H; source cited: Ansari et al. (2024). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions). Table 6 cites Ansari et al. (2024), i.e. the Chronos NREL solar data.
-
Taxi · split: unknown
Table 6 frequencies: 30min, 1H; source cited: Salinas et al. (2019). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Uber TLC · split: unknown
Table 6 frequencies: 1H, 1D; source cited: FiveThirtyEight (2025). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
USHCN · split: unknown
Table 6 frequencies: 1D, 1W; source cited: Ansari et al. (2024). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Weatherbench · split: unknown
Table 6 frequencies: 1H, 1D, 1W; source cited: Rasp et al. (2020). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Wiki · split: unknown
Table 6 frequencies: 1H, 1D, 1W; source cited: Ansari et al. (2024). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Wind Farms · split: unknown
Table 6 frequencies: 1H, 1D; source cited: Godahewa et al. (2021). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Temperature Rain · split: unknown
Table 6 frequencies: 1D; source cited: Godahewa et al. (2021). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
London Smart Meters · split: unknown
Table 6 frequencies: 30min, 1D; source cited: Godahewa et al. (2021). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Alibaba Cluster Trace (2018) · split: unknown
Table 6 frequencies: 5min, 1H; source cited: Woo et al. (2023). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Azure VM Traces (2017) · split: unknown
Table 6 frequencies: 5min, 1H; source cited: Woo et al. (2023). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Borg Cluster Data (2011) · split: unknown
Table 6 frequencies: 5min, 1H; source cited: Woo et al. (2023). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
LargeST (PeMS California) · split: unknown
Table 6 frequencies: 1H, 1D; source cited: Liu et al. (2023). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Q-Traffic · split: unknown
Table 6 frequencies: 15min, 1H; source cited: Jiang et al. (2023). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
Buildings 900K · split: unknown
Table 6 frequencies: 1H, 1D; source cited: Emami et al. (2023). Drawn from the Chronos and/or GIFT-Eval pretraining corpora; per the paper, GIFT-Eval test portions were excluded and partial overlap with GIFT-Eval training portions exists (see model-level exclusions).
-
TSI (Trend, Seasonality, and Irregularity) · split: unknown
Synthetic univariate generator.
Evidence
“TSI (Trend, Seasonality, and Irregularity): based on Bahrpeyma et al. (2021), this generator produces diverse synthetic series by randomly constructing and combining different trend, seasonality, and irregularity components.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
TCM (Temporal Causal Model) · split: unknown
Synthetic univariate generator.
Evidence
“TCM (Temporal Causal Model): this generator samples random causal graphs from a temporal causal model (Runge et al., 2023), from which time series are generated via autoregression.”
arxiv.org · paper · Section 4.1 · retrieved 2026-10-15 -
KernelSynth · split: unknown
Used as a base generator for the multivariatizers (multivariate/covariate synthetic tasks).
Evidence
“As base univariate generators, we employed a diverse set including autoregressive (AR) models, exponential smoothing (ETS) models, TSI, and KernelSynth (Ansari et al., 2024).”
arxiv.org · paper · Section 4.2 · retrieved 2026-10-15 -
Multivariatizers (cotemporaneous and sequential) over AR, ETS, TSI and KernelSynth base generators · split: unknown
Synthetic multivariate and covariate-informed tasks.
Evidence
“For multivariate and covariate-informed tasks, we relied entirely on synthetic data.”
arxiv.org · paper · Section 4.2 · retrieved 2026-10-15 “A multivariatizer samples multiple time series from base univariate generators and imposes dependencies among them to create multivariate dynamics.”
arxiv.org · paper · Section 4.2 · retrieved 2026-10-15
Declared exclusions
-
Benchmark GIFT-Eval · test portion excluded
Evidence
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
arxiv.org · paper · Section 5.1 (GIFT-Eval) · retrieved 2026-10-15 “Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
arxiv.org · paper · Section 5.1 (GIFT-Eval) · retrieved 2026-10-15 “Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
huggingface.co · model card · Section 'Training data' · retrieved 2026-10-15
Benchmark cells
GIFT-Eval
Not seen Jena Weather (GIFT-Eval) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen BizITObs - Application · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen BizITObs - Service · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen BizITObs - L2C · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Bitbrains - Fast Storage · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Bitbrains - rnd · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Restaurant (Recruit Restaurant Visitor Forecasting) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen ETT1 (GIFT-Eval) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen ETT2 (GIFT-Eval) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Loop Seattle (LibCity) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen SZ-Taxi (LibCity) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen M_DENSE (LibCity) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Solar (LSTNet, 137 Alabama PV plants) · related dataset NREL Solar Power Data for Integration Studies (2006) was used; the authors state this dataset's test portion was excluded
Evidence (7)
“Solar”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“The uploaded dataset contains the aggregated version of a subset of the original dataset used by Lai et al. (2017).”
“Solar (5 Min., Hourly) contains data about solar power generation in the US in 2006. The original data has 5 minute frequency and was obtained from https://www.nrel.gov/grid/solar-power-data.html”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Hierarchical Sales (Mancuso et al., 2020) · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen M4 Yearly · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen M4 Quarterly · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Training data seen M4 Monthly · the dataset was used; the source does not say which part; the test portion was explicitly excluded
Evidence (5)
“M4 (Monthly)”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Training data seen M4 Weekly · the dataset was used; the source does not say which part; the test portion was explicitly excluded
Evidence (5)
“M4 (Weekly)”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Training data seen M4 Daily · the dataset was used; the source does not say which part; the test portion was explicitly excluded
Evidence (5)
“M4 (Daily)”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Training data seen M4 Hourly · the dataset was used; the source does not say which part; the test portion was explicitly excluded
Evidence (5)
“M4 (Hourly)”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Hospital · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen COVID Deaths · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen US Births · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Saugeen River Flow · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Training data seen Temperature Rain · the dataset was used; the source does not say which part; the test portion was explicitly excluded
Evidence (5)
“Temperature-Rain”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Training data seen KDD Cup 2018 · the dataset was used; the source does not say which part; the test portion was explicitly excluded
Evidence (5)
“KDD Cup (2018)”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Not seen Car Parts · the authors state its test portion was excluded from training
Evidence (3)
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Training data seen ElectricityLoadDiagrams20112014 (UCI) · the dataset was used; the source does not say which part; the test portion was explicitly excluded
Evidence (5)
“Electricity”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“When constructing the pretraining corpus for Chronos-2, we carefully ensured that it did not overlap with the test portions of any GIFT-Eval task at any sampling frequency.”
“Nonetheless, the corpus does include partial overlap with the training portions of some GIFT-Eval datasets.”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
Long-horizon forecasting (ETT, Electricity, Traffic, Weather, Exchange, ILI)
Not seen ETTh1 · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen ETTh2 · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen ETTm1 · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen ETTm2 · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Electricity (ECL, 321 clients) · related dataset ElectricityLoadDiagrams20112014 (UCI) was used
Evidence (5)
“Electricity”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“The raw dataset is in https://archive.ics.uci.edu/ml/datasets/ElectricityLoadDiagrams20112014.”
“Because the some dimensions are equal to 0. So we eliminate the records in 2011. Final we get data contains electircity consumption of 321 clients from 2012 to 2014.”
“https://archive.ics.uci.edu/ml/datasets/ElectricityLoadDiagrams20112014”
Traffic (PeMS San Francisco Bay, 862 sensors) · related dataset LargeST (PeMS California) was used
Evidence (3)
“LargeST (2017)”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“The PeMS is a popular source of data for many traffic forecasting datasets such as PEMS03, PEMS04, PEMS07, PEMS08, and PEMS Bay, as well as the popular Traffic dataset from (Lai et al., 2018).”
Not seen Weather (Jena, 2020; Autoformer) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Exchange-Rate · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen ILI (CDC influenza-like illness, national) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Time-series anomaly detection
Not seen TSB-AD-U (TSB-AD univariate) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen TSB-AD-M (TSB-AD multivariate) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen UCR Time Series Anomaly Archive (Hexagon ML/UCR, KDD Cup 2021) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Server Machine Dataset (SMD) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Mars Science Laboratory (MSL) telemetry anomaly dataset · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Soil Moisture Active Passive (SMAP) telemetry anomaly dataset · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Pooled Server Metrics (PSM) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Secure Water Treatment (SWaT) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Yahoo S5 Labeled Anomaly Detection Dataset (Yahoo Webscope S5) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen Numenta Anomaly Benchmark (NAB) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Time-series classification
Not seen UCR Time Series Classification Archive (2018) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Not seen UEA Multivariate Time Series Classification Archive (2018) · absent from the complete list of training data
Evidence (6)
“We incorporated select datasets from the Chronos (Ansari et al., 2024) and GIFT-Eval (Aksu et al., 2024) pretraining corpora into Chronos-2’s training corpus.”
“The full list of datasets is provided in Table 6 (Appendix).”
“Table 6: Real univariate datasets used for pretraining Chronos-2.”
“Subset of [GIFT-Eval Pretrain](https://huggingface.co/datasets/Salesforce/GiftEvalPretrain)”
“Subset of [Chronos Datasets](https://huggingface.co/datasets/autogluon/chronos_datasets) (excluding test portion of datasets that overlap with GIFT-Eval)”
“Synthetic univariate and multivariate data”
Notes
120M-parameter main model; autogluon/chronos-2-small (28M) is listed here because its model card declares the same datasets metadata (autogluon/chronos_datasets, Salesforce/GiftEvalPretrain) and the paper describes it only as 'a small model with 28M parameters' (ablation, Section 5.4). Real data is univariate only (Table 6: 23 datasets, from the Chronos corpus and GIFT-Eval pretrain); all multivariate/covariate data is synthetic. Trained on GIFT-Eval pretrain data and partially on GIFT-Eval datasets' training portions; GIFT-Eval test portions explicitly excluded. Which subset of GIFT-Eval Pretrain and which time ranges were used is not specified. Chronos-2 is two-stage (2048 context pre-training, then 8192 context post-training); the data of each stage is not separated. Other stated exclusions, not tracked here: fev-bench (all); chronos-benchmark-ii (all).
Sources
- Chronos-2: From Univariate to Universal Forecasting (technical report, arXiv v1) · paper · retrieved 2026-10-15
- amazon/chronos-2 model card · model card · retrieved 2026-10-15
- autogluon/chronos-2-small model card · model card · retrieved 2026-10-15
- amazon-science/chronos-forecasting README · readme · retrieved 2026-10-15