Full papers
Submitted for peer review; accepted papers are presented as talks or posters during the workshop and appear on OpenReview. Non-archival, allowing future publication elsewhere.
Machine Learning/CORE Unranked
ICML 2026
ICML 2026 Workshop on Foundations of Deep Generative Models: Understanding Memorization, Generalization, and Reasoning
The ICML 2026 Workshop on Foundations of Deep Generative Models (FoGen) brings together researchers to investigate how deep generative models learn, memorize, generalize, and perform structured reasoning. The workshop focuses on theoretical, empirical, and application-driven questions around diffusion, flow-based, and autoregressive models, with emphasis on reliability, interpretability, and scientific use cases.
Key deadlines
Verified September 9, 2026Full paper
May 9, 2026
AoE
Workshop timeline
Submission and decisions
Full paperKey deadline
May 9, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Memorization and Generalization
Empirical and theoretical studies of memorizationGeneralizationRegime transitionsRoles of capacity, data complexity, and scaling
02
Reasoning and Compositionality
Compositional inferenceCausal inferenceStructured inferenceIn-context learningChain-of-thoughtMulti-step generation
03
Optimization and Inductive Bias
Learning dynamicsArchitectureImplicit regularizationShaping memorization, generalization, and reasoning behavior
04
Evaluation and Benchmarking
Metrics for distinguishing memorization from generalizationDiagnostic frameworksRobustness benchmarksPrivacy benchmarksExtrapolation benchmarks
05
Scientific Discovery with DGMs
Scientific machine learningHealthcareProtein designMolecular discoveryInterpretability and reasoning in scientific workflows
Policies worth checking twice
- Submission is double-blind; manuscripts must be anonymized.
- Papers must not exceed 8 pages excluding references and appendices.
- Camera-ready versions may be extended by up to 2 additional pages (max 10 pages) for single-column format or 1 additional page (max 9 pages) for two-column format.
- Accepted papers are non-archival and may be submitted elsewhere after the workshop.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026