📝 Publications

PartEdit: Fine‑Grained Image Editing using Pre‑Trained Diffusion Models
ACM SIGGRAPH, 2025
Aleksandar Cvejic, Abdelrahman Eldesokey, Peter Wonka
[PDF] [Code] [Project page]

NearID: Identity Representation Learning via Near-identity Distractors
arXiv, 2026 (Under Review)
Aleksandar Cvejic, Rameen Abdal, Abdelrahman Eldesokey, Bernard Ghanem, Peter Wonka
KAUST × Snap Research
[PDF] [Code] [Project page] [Video]

EditCLIP: Representation Learning for Image Editing
IEEE ICCV, 2025
Qian Wang, Aleksandar Cvejic, Abdelrahman Eldesokey, Peter Wonka
[PDF] [Code] [Project page]

Mind-the-Glitch: Visual Correspondence for Detecting Inconsistencies in Subject-Driven Generation
NeurIPS, 2025 (Spotlight)
Abdelrahman Eldesokey, Aleksandar Cvejic, Bernard Ghanem, Peter Wonka
[PDF] [Code] [Project page]

CriterAlign: Criterion-Centric Rationale Alignment for Code Preference Judging
arXiv, 2026 (Under Review)
Zhenyu Li, Aleksandar Cvejic, Zehui Chen, Peter Wonka
[PDF] [Code] [Project page]

Architect-Ant: Editable Automatic Furnishing of Architectural Floor Plans
Under Review, 2026
Fedor Rodionov, Aleksandar Cvejic, Michael Birsak, John Femiani, Peter Wonka
A vision-language model that furnishes floor plans as an editable coordinate DSL; introduces the AntPlan-270 dataset.