Outcome-Based RL Provably Leads Transformers to Reason, but Only With the Right Data

Ähnliche Einträge

• Provable Benefits of Complex Parameterizations for Structured State Space Models
  von: Ran-Milo, Yuval, et al.
  Veröffentlicht: (2024)
• Do Neural Networks Need Gradient Descent to Generalize? A Theoretical Study
  von: Alexander, Yotam, et al.
  Veröffentlicht: (2025)
• A Mechanistic Account of Attention Sinks in GPT-2: One Circuit, Broader Implications for Mitigation
  von: Ran-Milo, Yuval, et al.
  Veröffentlicht: (2026)
• Attention Sinks Are Provably Necessary in Softmax Transformers: Evidence from Trigger-Conditional Tasks
  von: Ran-Milo, Yuval
  Veröffentlicht: (2026)
• What Makes Data Suitable for a Locally Connected Neural Network? A Necessary and Sufficient Condition Based on Quantum Entanglement
  von: Alexander, Yotam, et al.
  Veröffentlicht: (2023)