GITCO: Gated Inference-Time Context Optimization in TSFMs

Patch-based Time Series Foundation Models (TSFMs) suffer from \textit{context poisoning:} structurally anomalous patches capture disproportionate attention and degrade zero-shot forecast quality. We propose improving TSFM accuracy at inference time by optimizing the input context rather than modifying model weights. We present Gated Inference-Time Context Optimization (GITCO), a three-component framework: \textit{Gate, Router and Critic} that selectively suppresses harmful patches. Evaluated on TimesFM 2.5 across 53 GIFT-Eval datasets under K-fold cross-validation, GITCO achieves an average +1.9\% MASE reduction on TimesFM 2.5 while capturing 89.9\% of the improvement upper bound. We introduce \textit{context sensitivity profiles} as a new characterizable property of TSFMs: the mapping from time series meta-features to expected accuracy improvement under inference-time context intervention, shaped jointly by model architecture and the statistical structure of the data.