How to Allocate, How to Learn? Dynamic Rollout Allocation and Advantage Modulation for Policy Optimization

Documenti analoghi

• MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning
  di: Fu, Xiaoliang, et al.
  Pubblicazione: (2026)
• Placing Puzzle Pieces Where They Matter: A Question Augmentation Framework for Reinforcement Learning
  di: Fang, Yangyi, et al.
  Pubblicazione: (2026)
• From $\log π$ to $π$: Taming Divergence in Soft Clipping via Bilateral Decoupled Decay of Probability Gradient Weight
  di: Fu, Xiaoliang, et al.
  Pubblicazione: (2026)
• Proximity-Based Multi-Turn Optimization: Practical Credit Assignment for LLM Agent Training
  di: Fang, Yangyi, et al.
  Pubblicazione: (2026)
• OptPO: Optimal Rollout Allocation for Test-time Policy Optimization
  di: Wang, Youkang, et al.
  Pubblicazione: (2025)