We invite novel contributions that relate broadly to self-improvement in foundation models, accepting submissions of both empirical and theoretical nature. Relevant topics include (but are not limited to): new learning objectives and algorithms, multi-agent and multi-model systems, machine-generated synthetic data, autonomous online learning and reinforcement learning algorithms, efficient exploitation of tools and external information, theoretically characterizing conditions under which self-improvement is feasible, weak supervision for improving strong models, inference time improvements, limits of self-improvement training, and applications in software agents, robotics, multi-modal systems, math, etc.
This year, ICLR is discontinuing the separate “Tiny Papers” track, and is instead requiring each workshop to accept short (up to 4 pages) paper submissions, with an eye towards inclusion; see https://iclr.cc/Conferences/2025/CallForTinyPapers for more details. Authors of these papers will be earmarked for potential funding from ICLR, but need to submit a separate application for Financial Assistance that evaluates their eligibility. This application for Financial Assistance to attend ICLR 2025 will become available on https://iclr.cc/Conferences/2025/ at the beginning of February and close on March 2nd.
Paper Submissions Due: February 11, 2025 (AOE)
Notification of Acceptance: March 5, 2025
Camera-ready Paper Due: TBD
Workshop Date: TBD
Submission URL: https://openreview.net/group?id=ICLR.cc/2025/Workshop/SSI-FM
Submission Instructions: Regular submissions can be up to 9 pages and tiny paper submissions can be up to 4 pages in ICLR format (double-blind) [format], excluding references and supplementary material. We allow an unlimited number of pages for references and supplementary material, but reviewers are not required to review the supplementary material. Accepted papers will be presented at the workshop as contributed talks and/or posters. At the discretion of authors, accepted papers can be published through the workshop website.
Concurrent Submissions: We welcome research papers currently under review at archival NLP and ML conferences (e.g., EMNLP, and ICML). Submission to this workshop will not break the anonymity or dual submission policies for these conferences. The workshop is non-archival. Please note that we do allow the submission of recently published work. However, when selecting papers for oral presentation, preference is given to original works.
Double-blind reviews: Submissions will be peer-reviewed by at least 2 reviewers, in addition to an area chair. The reviewing process will be double-blind at the level of the reviewers. As an author, you are responsible for anonymizing your submission. Do not include any authors' names, affiliations, acknowledgements, or any other information that could result in de-anonymization.