What Drives Compositional Generalization? The Importance of Continuous Training Objectives in Visual Generative Models

Documenti analoghi

• When and How Does CLIP Enable Domain and Compositional Generalization?
  di: Kempf, Elias, et al.
  Pubblicazione: (2025)
• Two Effects, One Trigger: On the Modality Gap, Object Bias, and Information Imbalance in Contrastive Vision-Language Models
  di: Schrodi, Simon, et al.
  Pubblicazione: (2024)
• Eureka-Moments in Transformers: Multi-Step Tasks Reveal Softmax Induced Optimization Problems
  di: Hoffmann, David T., et al.
  Pubblicazione: (2023)
• Concept Bottleneck Models Without Predefined Concepts
  di: Schrodi, Simon, et al.
  Pubblicazione: (2024)
• Orbis: Overcoming Challenges of Long-Horizon Prediction in Driving World Models
  di: Mousakhan, Arian, et al.
  Pubblicazione: (2025)