State Space Models are Provably Comparable to Transformers in Dynamic Token Selection

Documenti analoghi

• Mixture of Experts Provably Detect and Learn the Latent Cluster Structure in Gradient-Based Learning
  di: Kawata, Ryotaro, et al.
  Pubblicazione: (2025)
• Degrees of Freedom for Linear Attention: Distilling Softmax Attention with Optimal Feature Efficiency
  di: Nishikawa, Naoki, et al.
  Pubblicazione: (2025)
• Transformers Provably Solve Parity Efficiently with Chain of Thought
  di: Kim, Juno, et al.
  Pubblicazione: (2024)
• From Shortcut to Induction Head: How Data Diversity Shapes Algorithm Selection in Transformers
  di: Kawata, Ryotaro, et al.
  Pubblicazione: (2025)
• In-Context Learning Is Provably Bayesian Inference: A Generalization Theory for Meta-Learning
  di: Wakayama, Tomoya, et al.
  Pubblicazione: (2025)