Can Looped Transformers Learn to Implement Multi-step Gradient Descent for In-context Learning?

Documenti analoghi

• On the Role of Depth and Looping for In-Context Learning with Task Diversity
  di: Gatmiry, Khashayar, et al.
  Pubblicazione: (2024)
• Reasoning with Latent Thoughts: On the Power of Looped Transformers
  di: Saunshi, Nikunj, et al.
  Pubblicazione: (2025)
• On the Inductive Bias of Stacking Towards Improving Reasoning
  di: Saunshi, Nikunj, et al.
  Pubblicazione: (2024)
• Simplicity Bias via Global Convergence of Sharpness Minimization
  di: Gatmiry, Khashayar, et al.
  Pubblicazione: (2024)
• Landscape-Aware Growing: The Power of a Little LAG
  di: Karp, Stefani, et al.
  Pubblicazione: (2024)