Deep Continual Learning in the Foundation Model Era (Dagstuhl Seminar 25432)

Kanan, Christopher; Mundt, Martin; Tuytelaars, Tinne; van de Weijer, Joost; Hess, Timm Felix

doi:10.4230/DagRep.15.10.119

Document

Deep Continual Learning in the Foundation Model Era (Dagstuhl Seminar 25432)

Authors Christopher Kanan, Martin Mundt, Tinne Tuytelaars, Joost van de Weijer, Timm Felix Hess and all authors of the abstracts in this report

Part of: Issue: Dagstuhl Reports, Volume 15, Issue 10
Part of: Volume: Dagstuhl Reports, Volume 15
Part of: Journal: Dagstuhl Reports (DagRep)
License: Creative Commons Attribution 4.0 International license
Publication Date: 2026-06-23

PDF

File

PDF

DagRep.15.10.119.pdf

Filesize: 2 MB
16 pages

Document Identifiers

DOI: 10.4230/DagRep.15.10.119
URN: urn:nbn:de:0030-drops-254123

Subject Classification

ACM Subject Classification

Computing methodologies → Artificial intelligence
Computing methodologies → Neural networks
Computing methodologies → Learning settings
Computing methodologies → Learning paradigms
Computing methodologies → Bio-inspired approaches
Computing methodologies → Learning latent representations
Computing methodologies → Computer vision
Computing methodologies → Natural language processing
Computing methodologies → Distributed artificial intelligence

Keywords

continual learning
deep learning
foundation models

Metrics

Access Statistics
Total Accesses (updated on a weekly basis)

0

Document

0

Metadata

Abstract

This report documents the program and the outcomes of Dagstuhl Seminar 25432 "Deep Continual Learning in the Foundation Model Era". This seminar brought together 23 researchers to discuss research at the intersection of continual learning and foundation models. The discussion centered on the major challenges arising from continual training of foundation models, including the need for new benchmarks, new opportunities for memory-based continual learning, emerging application domains, and the development of efficient metrics to quantify forgetting of foundation model knowledge. In addition, the report contains a summery of the talks of the partipants.

Cite As Get BibTex

Christopher Kanan, Martin Mundt, Tinne Tuytelaars, Joost van de Weijer, and Timm Felix Hess. Deep Continual Learning in the Foundation Model Era (Dagstuhl Seminar 25432). In Dagstuhl Reports, Volume 15, Issue 10, pp. 119-134, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2026) https://doi.org/10.4230/DagRep.15.10.119

Author Details

Christopher Kanan

University of Rochester, US

Martin Mundt

Universität Bremen, DE

Tinne Tuytelaars

KU Leuven, BE

Joost van de Weijer

Computer Vision Center - Barcelona, ES

Timm Felix Hess

KU Leuven, BE

and all authors of the abstracts in this report

Any Issues?

Feedback on the Current Page

Thanks for your feedback!

Feedback submitted to Dagstuhl Publishing

Could not send message

Please try again later or send an E-mail