method · first seen 12 Nov 2026
pessimistic value iteration
Value iteration with uncertainty-based penalties that discourages overestimating poorly covered actions in offline settings.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Offline KL-Regularized Reinforcement Learning with Stochastic TransitionsPapers · 12 Nov 2026 · Andrew Liu
Latest sources
- Offline KL-Regularized Reinforcement Learning with Stochastic TransitionsPapers · 12 Nov 2026 · Andrew Liu