GITCO: Gated Inference-Time Context Optimization in TSFMs

2026-06-03 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

GITCO (Gated Inference-Time Context Optimization) is a new three-component framework designed to improve the accuracy of patch-based Time Series Foundation Models (TSFMs) by addressing context poisoning. This issue occurs when structurally anomalous patches disproportionately capture attention, silently degrading zero-shot forecast quality. GITCO, comprising a Gate, Router, and Critic, operates at inference time to selectively identify and suppress these harmful patches without requiring any modifications to the model's weights. Evaluated on TimesFM 2.5 across 53 GIFT-Eval datasets using K-fold cross-validation, GITCO achieved an average +1.95% MASE reduction on TimesFM 2.5, capturing 89.9% of the improvement upper bound. The authors also introduce context sensitivity profiles as a new characterizable property of TSFMs.

Key takeaway

For Machine Learning Engineers deploying Time Series Foundation Models, you should consider integrating inference-time context optimization techniques like GITCO. This approach offers a significant +1.95% MASE reduction on models like TimesFM 2.5 by mitigating context poisoning without retraining. Implementing GITCO allows you to enhance forecast quality and model robustness in production environments, particularly for zero-shot applications where data anomalies are common.

Key insights

Patch-based TSFMs can be improved at inference time by optimizing input context to suppress harmful patches.

Principles

• Context poisoning degrades TSFM zero-shot forecasts.
• Inference-time context optimization enhances TSFM accuracy.
• Model architecture and data structure shape context sensitivity.

Method

GITCO uses a Gate, Router, and Critic to identify and suppress structurally anomalous patches in TSFM input context during inference, without altering model weights.

In practice

• Apply GITCO to enhance TimesFM 2.5 forecast accuracy.
• Characterize TSFM context sensitivity profiles.

Topics

• Time Series Foundation Models
• Context Optimization
• Inference-Time Optimization
• Context Poisoning
• GITCO Framework
• TimesFM 2.5

Best for: Research Scientist, AI Scientist, Machine Learning Engineer

Related on AIssential

• GITCO: Gated Inference-Time Context Optimization in TSFMs — GITCO improves TSFM accuracy by optimizing input context at inference time, mitigating "context poisoning" without model weight updates.
• Five Questions About Chronos-2, the Time Series Foundation Model — Time series foundation models like Chronos-2 enable zero-shot forecasting by learning recurring temporal shapes from diverse data.
• Why Do Time Series Models Need Long Context Windows? — Long context windows in time series forecasting primarily reduce uncertainty in identifying the data-generating process, crucial for optimal predictions.
• Graph In-Context Operator Networks for Generalizable Spatiotemporal Prediction — In-context operator learning with diverse operators significantly outperforms classical methods for complex spatiotemporal prediction.
• Interventional Time Series Priors for Causal Foundation Models — CausalTimePrior generates synthetic interventional time series data to train foundation models for time series causal inference.
• The digital quant: instant portfolio optimization with JointFM — JointFM is a zero-shot foundation model predicting full multivariate distributions for instant, domain-agnostic time-series forecasting.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.