analiza aiqbench
This paper proposes a black-box correction layer for time series foundation models that combines retrieved historical analogues with the model's own past forecast errors, weighting the two by recent performance. The reported result is a consistent MSE reduction across 45 backbone-dataset pairs, averaging 9.26%, but the gains are measured on benchmark datasets against specific TSFMs, so practical value will depend on whether they hold in other domains and on how much the method costs to run. For ML practitioners, the main takeaway is that wrapping a frozen foundation model with lightweight, feedback-driven correction may be a cheaper alternative to fine-tuning.
ECHO: Correcting Time Series Foundation Model Forecasts with Historical Analogues and Errors
arXiv:2610.16433v1 Announce Type: new Abstract: Time Series Foundation Models (TSFMs) provide general-purpose forecasting capabilities, yet their forecasts can remain misaligned with the temporal characteristics of a target domain. This motivates lightweight forecast correction even when only black-box access to the TSFM is available. Our key observation is that two forms of historical information provide complementary correction signals: historical analogues reveal how similar contexts evolved, while past forecast errors reveal a TSFM's characteristic deviations in the target domain. Based on this observation, we propose ECHO, a black-box forecast correction framework that learns from both sources. ECHO uses AnalogE to assess retrieved historical analogues together with the current context and base forecast, learning transferable corrections through cross-domain pretraining. In parallel, RecentE learns target-specific residual corrections from past forecast errors and refines them as causally available feedback accumulates. ECHO dynamically combines the two corrections according to their observed forecasting performance. Across 5 representative TSFMs and 9 real-world datasets, ECHO reduces MSE in all 45 backbone--dataset settings, yielding an average relative reduction of 9.26\% in macro-average MSE. Its gains also extend to long-horizon forecasting.
Abstract:Time Series Foundation Models (TSFMs) provide general-purpose forecasting capabilities, yet their forecasts can remain misaligned with the temporal characteristics of a target domain. This motivates lightweight forecast correction even when only black-box access to the TSFM is available. Our key observation is that two forms of historical information provide complementary correction signals: historical analogues reveal how similar contexts evolved, while past forecast errors reveal a TSFM's characteristic deviations in the target domain. Based on this observation, we propose ECHO, a black-box forecast correction framework that learns from both sources. ECHO uses AnalogE to assess retrieved historical analogues together with the current context and base forecast, learning transferable corrections through cross-domain pretraining. In parallel, RecentE learns target-specific residual corrections from past forecast errors and refines them as causally available feedback accumulates. ECHO dynamically combines the two corrections according to their observed forecasting performance. Across 5 representative TSFMs and 9 real-world datasets, ECHO reduces MSE in all 45 backbone--dataset settings, yielding an average relative reduction of 9.26\% in macro-average MSE. Its gains also extend to long-horizon forecasting.
Machine Learning (cs.LG)
arXiv:2610.16433 [cs.LG]
(or arXiv:2610.16433v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.16433 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 269 cuvinte
Subiecte conexe
News
Quantum Computing Report · 25 iul. 2026
Teger: Spatiotemporal Covariance for Probabilistic Traffic Forecasting
arXiv · 6 oct. 2026
Presentation: Designing AI Platforms for Reliability: Tools for Certainty, Agents for Discovery
InfoQ · 27 mai 2026
FinVerse: Financial Time-Series Benchmark
arXiv · 5 aug. 2026