Submission (via OpenReview) is currently open!

Dynamical systems have played important roles in the analysis and design of algorithms, paving the way for establishing non-asymptotic convergence guarantees in optimization, sampling, and equilibrium computation in games, which in turn have driven the success of modern artificial intelligence systems. Yet, the distinct mathematical backbone of these tools often creates barriers to entry for researchers and practitioners in machine learning. This workshop aims to lower that barrier by highlighting the unifying role of dynamical systems across these domains. We will convene optimization, sampling, and game theory experts, along with students and junior researchers in these areas, to foster cross-disciplinary dialogue and collaboration. 

We invite contributions from researchers in the form of papers related to dynamics and algorithms, including the topics listed below. All accepted papers will be presented as posters or selected for contributed talks. There will be no proceedings; however, accepted papers will be made available through the OpenReview website and listed on this site. 

Double-blind & Dual submissions

• Submissions must be properly anonymized for double-blind review.

• Papers already accepted at venues with archival proceedings (including the NeurIPS main conference) will not be considered.

• We discourage dual submissions to multiple NeurIPS workshops—please submit to the one that best fits your work.

• Extended abstracts of papers under review at other conferences/journals can be submitted if this is ok for the conference/journal in question (if in doubt, please check with them first). 

Topics:

Topics include, but are not limited to, the following aspects:

• Diffusion-based modeling: Can insights from dynamical systems and numerical analysis improve training efficiency, prevent mode collapse, and enable a principled understanding of guidance for diffusion models?

• Sampling: How to enable efficient and projection-free Langevin-based algorithms for sampling from constrained distributions, e.g., arising from fairness, prior knowledge, or interactions with physical systems?

• Optimization: Can insights from dynamical systems guide the algorithm choice for pretraining or fine-tuning large language models? How can implicit regularization be achieved? What can we say about fixed-point computation?

• Distributed learning: How to use tools from dynamical systems to understand the interaction of distributed machine learning systems with each other, or with surrounding social, biological, physical, or economic systems?

• Games: How can dynamical concepts (e.g., Hamiltonian dynamics and symplectic geometry) be used to further refine the analysis of games? How to develop algorithms for computing more general Nash equilibria (e.g., characterized by quasi-variational inequalities or periodic solutions)? What are emerging machine learning applications in view of agentic AI systems?

• Applications: Which emerging applications should be targeted for analysis from a theoretical or fundamental dynamical point of view? What insights can we gain? What is the expected impact?

Important Dates

• Submission Deadline: August 22, 2025 (AoE)

• Acceptance Notification: September 22, 2025 (AoE)

• Camera Ready: November 25, 2025 (AoE)

• Workshop: December 6 or 7, 2025 

Submission Instructions

• Submission: https://openreview.net/group?id=NeurIPS.cc/2025/Workshop/DynaFront

• Page limit: Recommended length is 4 to 5 pages (excluding references and supplementary materials). You may use additional pages for supplementary materials as necessary, but please make sure the main content of the paper and sufficient details are presented in the first 4-5 pages.

• Formatting: The following template is required: zip-file or overleaf-link.

• Reviews: Submissions will be evaluated by a reviewing committee. There will be a single round of reviews and no author response.

Questions? Check out the FAQs or reach us at dynafront.neurips@gmail.com.