Contamination Map Methodology

Time series / Dataset

KDD Cup 2018

270 hourly air-quality series from 59 stations in Beijing and London, 2017-2018 (KDD Cup 2018 competition); GIFT-Eval uses it at H and D.

Also known as
kdd_cup_2018, kdd_cup_2018_with_missing, KDD Cup 2018
Domain
nature
License
CC BY 4.0 (license of the Monash repository Zenodo upload)
Homepage
zenodo.org

Exposure by model

BenchmarkModelLevelReason
GIFT-Eval Chronos-2-Synth Not seen absent from the complete list of training data
GIFT-Eval Chronos-2 Training data seen the dataset was used; the source does not say which part; the test portion was explicitly excluded
GIFT-Eval Chronos-Bolt Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established (via Chronos training corpus (real-world datasets))
GIFT-Eval Chronos (T5) Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established (via Chronos training corpus (real-world datasets))
GIFT-Eval DADA (General Time Series Anomaly Detector with Adaptive Bottlenecks and Dual Adversarial Decoders) Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval Granite-TimeSeries-TTM-R1 Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval Granite-TimeSeries-TTM-R2.1 Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval Granite-TimeSeries-TTM-R2 Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval Granite-TTM-R3 Not seen the authors state its test portion was excluded from training
GIFT-Eval Lag-Llama Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established
GIFT-Eval Mantis (Mantis-8M) Unknown training data not disclosed in enough detail
GIFT-Eval Mantis+ (MantisPlus) Not seen absent from the complete list of training data
GIFT-Eval MantisV2 Not seen absent from the complete list of training data
GIFT-Eval Moirai-1.0-R Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established (via LOTSA (Large-scale Open Time Series Archive))
GIFT-Eval Moirai-1.1-R Unknown training data not disclosed in enough detail
GIFT-Eval Moirai-2.0-R Training data seen only the training split was used, not the benchmark's test data; the test portion was explicitly excluded
GIFT-Eval Moirai-MoE-1.0-R Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established (via LOTSA (Large-scale Open Time Series Archive))
GIFT-Eval MOMENT-1 Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established (via Time Series Pile)
GIFT-Eval NuTime Not seen absent from the complete list of training data
GIFT-Eval Sundial (Timer 3.0) Seen, split unclear the dataset was used; the source does not say which part (via Unified Time Series Dataset (UTSD)); this contradicts a statement that the data was excluded
GIFT-Eval Time-MoE Seen, split unclear the dataset was used; the source does not say which part (via Time-300B)
GIFT-Eval Time-RCD (TimeRCD) Not seen absent from the complete list of training data
GIFT-Eval Timer-S1 Training data seen a training split was used; its overlap with this benchmark's test data is not established (via LOTSA (Large-scale Open Time Series Archive)); the test portion was explicitly excluded
GIFT-Eval Timer-XL (timer-base-84m) Seen, split unclear the dataset was used; the source does not say which part (via Unified Time Series Dataset (UTSD))
GIFT-Eval Timer Seen, split unclear the dataset was used; the source does not say which part (via Unified Time Series Dataset (UTSD))
GIFT-Eval TimeRadar Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval TimesFM 1.0 Unknown training data not disclosed in enough detail
GIFT-Eval TimesFM 2.0 Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established (via LOTSA (Large-scale Open Time Series Archive))
GIFT-Eval TimesFM 2.5 Unknown training data not disclosed in enough detail
GIFT-Eval TimesFM 3.0 Unknown training data not disclosed in enough detail
GIFT-Eval TiRex 1.1 (gifteval) Not seen the authors state its test portion was excluded from training
GIFT-Eval TiRex-2-fevbench (TiRex-2-f) Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval TiRex-2-gifteval-pretrain (TiRex-2-gp) Not seen the authors state the dataset was excluded from training
GIFT-Eval TiRex-2-gifteval-zs (TiRex-2-g) Not seen the authors state the dataset was excluded from training
GIFT-Eval TiRex-2 Unknown training data not disclosed in enough detail
GIFT-Eval TiRex Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval Toto-Open-Base-1.0 Not seen the authors state its test portion was excluded from training
GIFT-Eval Toto-2.0-2.5B-FT Training data seen only the training split was used, not the benchmark's test data; the test portion was explicitly excluded
GIFT-Eval Toto 2.0 Not seen the authors state the dataset was excluded from training
GIFT-Eval Toto (2024 technical report) Seen, split unclear a training split was used; its overlap with this benchmark's test data is not established (via LOTSA (Large-scale Open Time Series Archive))
GIFT-Eval TSPulse r1 Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval TTM-R3 Training data seen only the training split was used, not the benchmark's test data; the test portion was explicitly excluded
GIFT-Eval Tiny Time Mixer (TTM) Research-Use R2 Seen, split unclear the dataset was used; the source does not say which part
GIFT-Eval UniTS few-shot anomaly detection checkpoints Not seen absent from the complete list of training data and from its base models' data
GIFT-Eval UniTS Not seen absent from the complete list of training data

Notes

Different from 'KDD Cup 2022' (wind power), which is in GIFT-Eval's pre-training split.

Sources