How JEPA Avoids Noisy Features: The Implicit Bias of Deep Linear Self Distillation Networks

Documenti analoghi

• Vanishing Gradients in Reinforcement Finetuning of Language Models
  di: Razin, Noam, et al.
  Pubblicazione: (2023)
• Text-Conditional JEPA for Learning Semantically Rich Visual Representations
  di: Huang, Chen, et al.
  Pubblicazione: (2026)
• Rethinking JEPA: Compute-Efficient Video SSL with Frozen Teachers
  di: Li, Xianhang, et al.
  Pubblicazione: (2025)
• Annotations Mitigate Post-Training Mode Collapse
  di: Springer, Jacob Mitchell, et al.
  Pubblicazione: (2026)
• Enhancing JEPAs with Spatial Conditioning: Robust and Efficient Representation Learning
  di: Littwin, Etai, et al.
  Pubblicazione: (2024)