Pré-Publication, Document De Travail Année : 2025

Are Time Series Foundation Models Susceptible to Catastrophic Forgetting?

Résumé

Time Series Foundation Models (TSFMs) have shown promising zero-shot generalization across diverse forecasting tasks. However, their robustness to continual adaptation remains underexplored. In this work, we investigate the extent to which TSFMs suffer from catastrophic forgetting when fine-tuned sequentially on multiple datasets. Using synthetic datasets designed with varying degrees of periodic structure, we measure the trade-off between adaptation to new data and retention of prior knowledge. Our experiments reveal that, while fine-tuning improves performance on new tasks, it often causes significant degradation on previously learned ones, illustrating a fundamental stability–plasticity dilemma.

Fichier principal
Vignette du fichier
Final_Version_CF.pdf (278.69 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05306471 , version 1 (09-10-2025)

Licence

Identifiants

Citer

Nouha Karaouli, Denis Coquenet, Elisa Fromont, Martial Mermillod, Marina Reyboz. Are Time Series Foundation Models Susceptible to Catastrophic Forgetting?. 2025. ⟨hal-05306471⟩
126 Consultations
150 Téléchargements

Altmetric

Partager

  • More