Short Papers
Four pages excluding references and appendices; presented as posters with selected submissions invited for talks. Must be submitted as a single PDF via OpenReview, double-blind reviewed.
Machine Learning/CORE Unranked
RPS Workshop 2026
Representations for the Physical Sciences Workshop @ NeurIPS 2026
The RPS Workshop 2026 is a NeurIPS satellite event focused on representation learning for physical sciences, addressing challenges in self-supervision, transfer learning, sampling, and tokenization of scientific data. It aims to foster discussions on developing representations that respect physical constraints like conservation laws, geometry, and causality, with applications in physics-based simulation, computational biology, and Earth system modeling.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research Notes
Accessible explanations of open problems, paradigms, datasets, or algorithms; no page limit; new results not required; submission channel to be announced.
Research areas in scope
01
Self-Supervision
Self-supervised learning for unlabeled scientific dataPreserving physical constraints in pretext tasksSemantic structures discovered under physical priors
02
Transfer
Extrapolation into physically meaningful regimesFalsifiability and actionability of learned embeddingsLimits of scientific transfer beyond training distributions
03
Sampling
Adaptive simulation for data generationClosed-loop data generation using simulatorsShaping training distributions via molecular dynamics, DFT, CFD, and PDE solvers
04
Tokenization
Discrete tokenization of continuous physical dataPreserving geometric priors and symmetriesNon-Euclidean and multi-scale tokenization methods
Policies worth checking twice
- Submissions are double-blind reviewed.
- Short Papers are limited to four pages of main content; references do not count toward the limit.
- Appendices are allowed but reviewers are not obliged to read them.
- Submissions must be a single PDF file; no additional files permitted.
- Authors may link to anonymized code or data repositories.
- Concurrent submissions to other venues are allowed, but authors must comply with those venues' dual-submission policies.
- The NeurIPS paper checklist is not required for this workshop.
- Program Committee members must follow the NeurIPS conflict-of-interest policy.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Official workshop siteworkshop LaTeX template ↓Privacy noticeR / PS Representations for Physical Sciences
Last verified September 9, 2026