Overcoming Reward Overoptimization via Adversarial Policy Optimization with Lightweight Uncertainty Estimation

Documenti analoghi

• Provably Mitigating Overoptimization in RLHF: Your SFT Loss is Implicitly an Adversarial Regularizer
  di: Liu, Zhihan, et al.
  Pubblicazione: (2024)
• Reviving The Classics: Active Reward Modeling in Large Language Model Alignment
  di: Shen, Yunyi, et al.
  Pubblicazione: (2025)
• Rethinking Reward Model Evaluation Through the Lens of Reward Overoptimization
  di: Kim, Sunghwan, et al.
  Pubblicazione: (2025)
• Iterative Data Smoothing: Mitigating Reward Overfitting and Overoptimization in RLHF
  di: Zhu, Banghua, et al.
  Pubblicazione: (2024)
• Scaling Laws for Reward Model Overoptimization in Direct Alignment Algorithms
  di: Rafailov, Rafael, et al.
  Pubblicazione: (2024)