CFP: DeLTa 2026 @ ICLR — Deep Generative Models (Theory, Principle, and Efficacy)

35 views

Skip to first unread message

Wei Huang

unread,

Feb 6, 2026, 8:10:52 PM (3 days ago) Feb 6

to Machine Learning News

Dear all,

We’re excited to announce the Call for Papers for the 2nd Workshop on Deep Generative Models in Machine Learning: Theory, Principle, and Efficacy (DeLTa 2026), to be held at ICLR 2026 (Rio de Janeiro, Brazil).

Important dates

Submission deadline: Feb 8, 2026
Notification: Mar 1, 2026
Camera-ready: Apr 5, 2026
Workshop date: TBD (ICLR 2026)

Scope (selected topics)

Unified theories bridging diffusion / flow-matching / energy-based / autoregressive models
Optimization, convergence, discretization, and noise schedules in diffusion & flow-matching
Stochastic control and optimal transport perspectives
Post-training theory (alignment, preference optimization, stability)
Implicit bias / regularization; information-theoretic and probabilistic analysis
Geometry & manifold learning for generative modeling
Algorithms & applications: LLM diffusion, one/few-step generation, multimodal generation, structured domains (graphs/meshes/manifolds), and AI4Science DGMs

Submission

Submission site: https://openreview.net/group?id=ICLR.cc/2026/Workshop/DeLTa
Double-blind; DeLTa workshop style template (https://delta-workshop.github.io/DeLTa2026/)
Short papers: up to 4 pages (excluding references/appendix)
Long papers: up to 8 pages (excluding references/appendix)
Non-proceedings: accepted papers appear on OpenReview; authors remain free to publish elsewhere

Short paper inclusion / ICLR support
ICLR has shifted workshop “tiny papers” into workshop-managed short-paper submissions; eligible short-paper authors may be considered for ICLR support (see ICLR guidance + financial assistance application).

If this aligns with your work, we’d love to see your submission and would really appreciate it if you could share this CFP with interested colleagues and students.