GIFT-Eval

The 'GIFT-Eval' dataset/benchmark is used to evaluate the performance of time series foundation models by providing a standardized set of tasks and metrics for assessing their effectiveness in learning transferable representations across diverse temporal patterns.

Papers using GIFT-Eval (20)

Chronos-2: From Univariate to Universal Forecasting2025 · 71 citesTempoPFN: Synthetic Pre-training of Linear RNNs for Zero-shot Time Series Forecasting2025 · 5 citesFrom Tables to Time: Extending TabPFN-v2 to Time Series Forecasting2025 · 4 citesAME-TS: Anchored Mixture-of-Experts for Time Series Forecasting2026Falcon-X: A Time Series Foundation Model for Heterogeneous Multivariate Modeling2026Kairos: Toward Adaptive and Parameter-Efficient Time Series Foundation Models2025 · 2 citesZero-shot Forecasting by Simulation Alone2026 · 1 citesReasoning-Aware Training for Time Series Forecasting2026Olivia: Harmonizing Time Series Foundation Models with Power Spectral Density2026Toto 2.0: Time Series Forecasting Enters the Scaling Era2026Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scaling2026A Foundation Model for Instruction-Conditioned In-Context Time Series Tasks2026EIDOS: Latent-Space Predictive Learning for Time Series Foundation Models2026Cisco Time Series Model Technical Report2025Xihe: Scalable Zero-Shot Time Series Learner Via Hierarchical Interleaved Block Attention2025TimeCopilot2025One-Embedding-Fits-All: Efficient Zero-Shot Time Series Forecasting by a Model Zoo2025VisionTS++: Cross-Modal Time Series Foundation Model with Continual Pre-trained Vision Backbones2025Output Scaling: YingLong-Delayed Chain of Thought in a Large Pretrained Time Series Forecasting Model2025TiRex: Zero-Shot Forecasting Across Long and Short Horizons with Enhanced In-Context Learning2025