6 papers
CriterAlign: Criterion-Centric Rationale Alignment for Code Preference Judging
Zhenyu Li, Aleksandar Cvejic, Zehui Chen +1
Pairwise human preference prediction is central to evaluating code-generation systems, where quality often depends on task-specific trade-offs beyond functional correctness. While…
Architect-Ant: Editable Automatic Furnishing of Architectural Floor Plans
Fedor Rodionov, Aleksandar Cvejic, Michael Birsak +2
Furnished floor plans are fundamental to real estate visualization, interior design, and architectural workflows. However, progress in automatic furniture arrangement has been limi…
NearID: Identity Representation Learning via Near-identity Distractors
Aleksandar Cvejic, Rameen Abdal, Abdelrahman Eldesokey +2
When evaluating identity-focused tasks such as personalized generation and image editing, existing vision encoders entangle object identity with background context, leading to unre…
Mind-the-Glitch: Visual Correspondence for Detecting Inconsistencies in Subject-Driven Generation
Abdelrahman Eldesokey, Aleksandar Cvejic, Bernard Ghanem +1
We propose a novel approach for disentangling visual and semantic features from the backbones of pre-trained diffusion models, enabling visual correspondence in a manner analogous…
PartEdit: Fine-Grained Image Editing using Pre-Trained Diffusion Models
Aleksandar Cvejic, Abdelrahman Eldesokey, Peter Wonka
We present the first text-based image editing approach for object parts based on pre-trained diffusion models. Diffusion-based image editing approaches capitalized on the deep unde…
EditCLIP: Representation Learning for Image Editing
Qian Wang, Aleksandar Cvejic, Abdelrahman Eldesokey +1
We introduce EditCLIP, a novel representation-learning approach for image editing. Our method learns a unified representation of edits by jointly encoding an input image and its ed…