Transformers Trained via Gradient Descent Can Provably Learn a Class of Teacher Models

Documenti analoghi

• Transformers Efficiently Perform In-Context Logistic Regression via Normalized Gradient Descent
  di: Zhang, Chenyang, et al.
  Pubblicazione: (2026)
• Gradient Descent Robustly Learns the Intrinsic Dimension of Data in Training Convolutional Neural Networks
  di: Zhang, Chenyang, et al.
  Pubblicazione: (2025)
• Looped Transformers with Layer Normalization Provably Learn the Power Method
  di: Wu, Lyumin, et al.
  Pubblicazione: (2026)
• Towards Simple and Provable Parameter-Free Adaptive Gradient Methods
  di: Tao, Yuanzhe, et al.
  Pubblicazione: (2024)
• CoPS: Empowering LLM Agents with Provable Cross-Task Experience Sharing
  di: Yang, Chen, et al.
  Pubblicazione: (2024)