ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Open questions and conjectures. (1) Generalization across n and d. Can a single conditional model, trained on a mixture of ball counts and dimensions, extrapolate to larger instances while maintaining or improving minimal separations after projection? (2) From finite boxes to periodic packings. • Projection-corrected sampling (PCFM): During sampling we interleave predict \to project \to correct. Each ODE step is projected onto the non- overlap constraints (pairwise and wall), followed by a small proximal re- laxation toward the learned flow to avoid fighting the projection. • Final refinement: A short physics-push/SRP stage removes residual con- tacts and aligns candidates with nearby local optima. Outlook. Even when not producing ultimate records, the sampler supplies high- quality, diverse seeds that substantially reduce time-to-solution for downstream local optimization. We view this as a construction heuristic that complements analytic bounds and may reveal structural patterns worth formalizing. Acknowledgements. I thank the organizers and participants for a stimulating workshop, and my collaborators Adam Zsolt Wagner, Jonas Klüver and Baran Hashemi for many contributions.

Record

Comments

Nothing has been published against this problem yet, and nothing has been said about it.

Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.

Sign in

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.