Metastable Dynamics of Chain-of-Thought Reasoning: Provable Benefits of Search, RL and Distillation

Documenti analoghi

• Transformers Provably Solve Parity Efficiently with Chain of Thought
  di: Kim, Juno, et al.
  Pubblicazione: (2024)
• Transformers Learn Nonlinear Features In Context: Nonconvex Mean-field Dynamics on the Attention Landscape
  di: Kim, Juno, et al.
  Pubblicazione: (2024)
• Transformers Provably Learn Chain-of-Thought Reasoning with Length Generalization
  di: Huang, Yu, et al.
  Pubblicazione: (2025)
• Drop the Act: Probe-Filtered RL for Faithful Chain-of-Thought Reasoning
  di: Parekh, Swapnil
  Pubblicazione: (2026)
• A Hierarchical Language Model with Predictable Scaling Laws and Provable Benefits of Reasoning
  di: Gaitonde, Jason, et al.
  Pubblicazione: (2026)