Forecasting on ETTm2 (MSE, MAE)
0.137MSERIDDE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RIDDEZero-shot=true, Input length=512, Forecasting horizon=64, Backbone=Chronos-Bolt2026.05 | 0.137 | 0.217 | |
| Cross-RAGZero-shot=true, Input length=512, Forecasting horizon=64, Backbone=Chronos-Bolt2026.05 | 0.146 | 0.225 | |
| RAF-Chronos-boltZero-shot=true, Input length=512, Forecasting horizon=64, Backbone=Chronos-Bolt2026.05 | 0.148 | 0.228 | |
| TS-RAGZero-shot=true, Input length=512, Forecasting horizon=64, Backbone=Chronos-Bolt2026.05 | 0.151 | 0.228 | |
| TimeBlocksEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.242 | — | |
| TimeBlocksEvaluation Protocol=Continual Forecasting, Coreset Strategy=StreamCore 5%, Calibration Steps=102026.06 | 0.244 | — | |
| TimeBlocksEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.249 | — | |
| ZEUSProtocol=Zero-shot, Model Category=Time Series Foundation Model2026.07 | 0.249 | 0.305 | |
| KairoslProtocol=Zero-shot, Model Category=Pretrained Forecasting Model2026.07 | 0.252 | 0.303 | |
| SundiallProtocol=Zero-shot, Model Category=Pretrained Forecasting Model2026.07 | 0.254 | 0.315 | |
| PatchTSTProtocol=Supervised, Model Category=Task-Specific Model2026.07 | 0.256 | 0.312 | |
| Time-MoElProtocol=Zero-shot, Model Category=Pretrained Forecasting Model2026.07 | 0.258 | 0.315 | |
| SymTimeEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.265 | — | |
| ModernTCNProtocol=Supervised, Model Category=Task-Specific Model2026.07 | 0.265 | 0.322 | |
| TotobProtocol=Zero-shot, Model Category=Pretrained Forecasting Model2026.07 | 0.267 | 0.303 | |
| GPT4TSProtocol=Supervised, Model Category=Task-Specific Model2026.07 | 0.275 | 0.331 | |
| TimeMixerEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.278 | — | |
| ChronosBoltbProtocol=Zero-shot, Model Category=Pretrained Forecasting Model2026.07 | 0.278 | 0.307 | |
| PatchTSTEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.28 | — | |
| TTMEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.28 | — | |
| Lag-LlamaEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.283 | — | |
| TimesNetProtocol=Supervised, Model Category=Task-Specific Model2026.07 | 0.286 | 0.328 | |
| TimesNetEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.289 | — | |
| PatchTSTEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.293 | — | |
| TimeMixerEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.302 | — | |
| TTMEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.302 | — | |
| UniTSProtocol=Zero-shot, Model Category=Time Series Foundation Model2026.07 | 0.328 | 0.362 | |
| TimesNetEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.33 | — | |
| AutoformerEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.334 | — | |
| Lag-LlamaEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.345 | — | |
| MOMENTProtocol=Zero-shot, Model Category=Time Series Foundation Model2026.07 | 0.359 | 0.388 | |
| TimerProtocol=Zero-shot, Model Category=Time Series Foundation Model2026.07 | 0.373 | 0.388 | |
| AutoformerEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.502 | — | |
| MomentEvaluation Protocol=Continual Forecasting, Coreset Strategy=Random 5%, Calibration Steps=102026.06 | 0.701 | — | |
| MomentEvaluation Protocol=Inference-time (no fine-tuning), Prediction Horizons=96, 192, 336, 7202026.06 | 0.71 | — |