A Theory of How Pretraining Shapes Inductive Bias in Fine-Tuning

Documenti analoghi

• Optimal Representation Size: High-Dimensional Analysis of Pretraining and Linear Probing
  di: Njaradi, Valentina, et al.
  Pubblicazione: (2026)
• A Theory of Initialisation's Impact on Specialisation
  di: Jarvis, Devon, et al.
  Pubblicazione: (2025)
• From Lazy to Rich: Exact Learning Dynamics in Deep Linear Networks
  di: Dominé, Clémentine C. J., et al.
  Pubblicazione: (2024)
• High-dimensional Analysis of Synthetic Data Selection
  di: Rezaei, Parham, et al.
  Pubblicazione: (2025)
• Inductive biases of multi-task learning and finetuning: multiple regimes of feature reuse
  di: Lippl, Samuel, et al.
  Pubblicazione: (2023)