Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models

Documenti analoghi

• On the Emergence of Linear Analogies in Word Embeddings
  di: Korchinski, Daniel J., et al.
  Pubblicazione: (2025)
• Symmetry in language statistics shapes the geometry of model representations
  di: Karkada, Dhruva, et al.
  Pubblicazione: (2026)
• Alternating Gradient Flows: A Theory of Feature Learning in Two-layer Neural Networks
  di: Kunin, Daniel, et al.
  Pubblicazione: (2025)
• Beyond Linear Response: Equivalence between Thermodynamic Geometry and Optimal Transport
  di: Zhong, Adrianne, et al.
  Pubblicazione: (2024)
• The lazy (NTK) and rich ($μ$P) regimes: a gentle tutorial
  di: Karkada, Dhruva
  Pubblicazione: (2024)