GRAM: Correcting Frozen Time-Series Foundation Models via Graph-Retrieved Amplitude Memory
cs.LG
Submitted: 2026-10-04
Updated: 2026-10-04
Code: https://github.com/ecnuabmlab/gram
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
The gist: Time-series foundation models (TSFMs) enable zero-shot forecasting through large-scale cross-domain pretraining, while retrieval augmentation further improves their performance by leveraging
Terminology
Abstract
Time-series foundation models (TSFMs) enable zero-shot forecasting through large-scale cross-domain pretraining, while retrieval augmentation further improves their performance by leveraging historical information. However, existing methods typically correct TSFM forecasts using the ground-truth futures of similar historical windows, which contain both predictive components already captured by the foundation model and sample-specific random fluctuation that is difficult to transfer. In contrast, recurring systematic model bias within prediction errors more directly characterizes the failure modes of a frozen TSFM and therefore provides more valuable correction signals. Effectively exploiting such model bias, however, poses two challenges: prediction errors at different numerical levels are difficult to compare due to scale differences, and the recurring bias must be extracted from prediction errors contaminated by random fluctuation. To address these challenges, we propose GRAM, a general retrieval-augmented framework for frozen TSFMs. GRAM first introduces an Amplitude Memory Module (AMM) that scales prediction errors by amplitude and aggregates them into retrievable prototypes. It then employs a Prototype Graph Module (PGM) to model relations among prototypes to aggregate consistent bias information while suppressing random fluctuation. During online forecasting, GRAM retrieves and expands prototypes relevant to the current query and generates per-horizon corrections to refine the original TSFM forecast. Experiments across multiple datasets and foundation models demonstrate consistent forecasting improvements.
Sources
- GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluation
- DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks
- Retrieval-Augmented Generation for Large Language Models: A Survey
- TimeGPT-1
- REALM: Retrieval-Augmented Language Model Pre-Training
- Atlas: Few-shot Learning with Retrieval Augmented Language Models
- Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting
- Moirai-MoE: Empowering Time Series Foundation Models with Sparse Mixture of Experts
- Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecasting
- Transformers in Time Series: A Survey
- TimeRAG: BOOSTING LLM Time Series Forecasting via Retrieval-Augmented Generation
- Stationarity-Aware Retrieval-Augmented Time Series Forecasting
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks