Imagen de Portada

Compositional Generalization Requires Linear, Orthogonal Representations in Vision Embedding Models

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Uselis, Arnas, Dittadi, Andrea, Oh, Seong Joon
Formato:	Preprint
Publicado:	2026
Materias:	Computer Vision and Pattern Recognition Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

CLIP Behaves like a Bag-of-Words Model Cross-modally but not Uni-modally
por: Koishigarina, Darina, et al.
Publicado: (2025)

How can embedding models bind concepts?
por: Uselis, Arnas, et al.
Publicado: (2026)

Does Data Scaling Lead to Visual Compositional Generalization?
por: Uselis, Arnas, et al.
Publicado: (2025)

Half-Truths Break Similarity-Based Retrieval
por: Kargi, Bora, et al.
Publicado: (2026)

On the rankability of visual embeddings
por: Sonthalia, Ankit, et al.
Publicado: (2025)

Diffusion Classifiers Understand Compositionality, but Conditions Apply
por: Jeong, Yujin, et al.
Publicado: (2025)

Sparse Autoencoders enable Robust and Interpretable Fine-tuning of CLIP models
por: Morelli, Fabian, et al.
Publicado: (2026)

When Do Diffusion Models learn to Generate Multiple Objects?
por: Jeong, Yujin, et al.
Publicado: (2026)

Intermediate Layer Classifiers for OOD generalization
por: Uselis, Arnas, et al.
Publicado: (2025)

Are Object-Centric Representations Better At Compositional Generalization?
por: Kapl, Ferdinand, et al.
Publicado: (2026)

Exploring the Effectiveness of Object-Centric Representations in Visual Question Answering: Comparative Insights with Foundation Models
por: Mamaghan, Amir Mohammad Karimi, et al.
Publicado: (2024)

Breaking the Likelihood-Quality Trade-off in Diffusion Models by Merging Pretrained Experts
por: Esfandiari, Yasin, et al.
Publicado: (2025)

Pretrained Visual Uncertainties
por: Kirchhof, Michael, et al.
Publicado: (2024)

Scalable Ensemble Diversification for OOD Generalization and Detection
por: Rubinstein, Alexander, et al.
Publicado: (2024)

DiffEnc: Variational Diffusion with a Learned Encoder
por: Nielsen, Beatrix M. G., et al.
Publicado: (2023)

MEME: Multi-entity & Evolving Memory Evaluation
por: Jung, Seokwon, et al.
Publicado: (2026)

Universal Algorithm-Implicit Learning
por: Woerner, Stefano, et al.
Publicado: (2026)

Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models
por: Berasi, Davide, et al.
Publicado: (2025)

Are We Done with Object-Centric Learning?
por: Rubinstein, Alexander, et al.
Publicado: (2025)

SCOPE: Semantic Coreset with Orthogonal Projection Embeddings for Federated learning
por: Hossen, Md Anwar, et al.
Publicado: (2026)

Multilingual Diversity Improves Vision-Language Representations
por: Nguyen, Thao, et al.
Publicado: (2024)

Mitigating Shortcut Learning with Diffusion Counterfactuals and Diverse Ensembles
por: Scimeca, Luca, et al.
Publicado: (2023)

Assessing Neural Network Robustness via Adversarial Pivotal Tuning
por: Christensen, Peter Ebert, et al.
Publicado: (2022)

Neighborhood-Adaptive Generalized Linear Graph Embedding with Latent Pattern Mining
por: Peng, S., et al.
Publicado: (2025)

Automated Learning of Semantic Embedding Representations for Diffusion Models
por: Jiang, Limai, et al.
Publicado: (2025)

Conceptualizing Embeddings: Sparse Disentanglement for Vision-Language Models
por: Kubaty, Piotr, et al.
Publicado: (2026)

CORE-MTL: Rethinking Gradient Balancing via Causal Orthogonal Representations
por: Wu, Chengfeng, et al.
Publicado: (2026)

Vision Foundation Model Embedding-Based Semantic Anomaly Detection
por: Ronecker, Max Peter, et al.
Publicado: (2025)

Training Without Orthogonalization, Inference With SVD: A Gradient Analysis of Rotation Representations
por: Choy, Chris
Publicado: (2026)

Linear Spaces of Meanings: Compositional Structures in Vision-Language Models
por: Trager, Matthew, et al.
Publicado: (2023)

Scratching Visual Transformer's Back with Uniform Attention
por: Hyeon-Woo, Nam, et al.
Publicado: (2022)

Learning Structured Representations with Hyperbolic Embeddings
por: Sinha, Aditya, et al.
Publicado: (2024)

Generation is Required for Data-Efficient Perception
por: Brady, Jack, et al.
Publicado: (2025)

Probing the Representational Power of Sparse Autoencoders in Vision Models
por: Olson, Matthew Lyle, et al.
Publicado: (2025)

Hier-COS: Making Deep Features Hierarchy-aware via Composition of Orthogonal Subspaces
por: Sani, Depanshu, et al.
Publicado: (2025)

Rotary Position Embedding for Vision Transformer
por: Heo, Byeongho, et al.
Publicado: (2024)

UniVL: Unified Vision-Language Embedding for Spatially Grounded Contextual Image Generation
por: Wang, Jiayun, et al.
Publicado: (2026)

Taming Feed-forward Reconstruction Models as Latent Encoders for 3D Generative Models
por: Wizadwongsa, Suttisak, et al.
Publicado: (2024)

SITUATE: Indoor Human Trajectory Prediction through Geometric Features and Self-Supervised Vision Representation
por: Capogrosso, Luigi, et al.
Publicado: (2024)

PHyCLIP: $\ell_1$-Product of Hyperbolic Factors Unifies Hierarchy and Compositionality in Vision-Language Representation Learning
por: Yoshikawa, Daiki, et al.
Publicado: (2025)