Expected Return Causes Outcome-Level Mode Collapse in Reinforcement Learning and How to Fix It with Inverse Probability Scaling

Documenti analoghi

• How does My Model Fail? Automatic Identification and Interpretation of Physical Plausibility Failure Modes with Matryoshka Transcoders
  di: Tang, Yiming, et al.
  Pubblicazione: (2025)
• Beyond Expected Return: Accounting for Policy Reproducibility when Evaluating Reinforcement Learning Algorithms
  di: Flageat, Manon, et al.
  Pubblicazione: (2023)
• Representation Collapsing Problems in Vector Quantization
  di: Zhao, Wenhao, et al.
  Pubblicazione: (2024)
• Multi-objective Reinforcement Learning with Nonlinear Preferences: Provable Approximation for Maximizing Expected Scalarized Return
  di: Peng, Nianli, et al.
  Pubblicazione: (2023)
• SOLAR: A Self-Optimizing Open-Ended Autonomous Agent for Lifelong Learning and Continual Adaptation
  di: Vetcha, Nitin, et al.
  Pubblicazione: (2026)