SCPO is a sampling-based weight-space projection algorithm that enforces rollout-evaluated safety constraints in policy optimization and provides a safe-by-induction guarantee from any safe starting point.
Safe Reinforcement Learning via Projection on a Safe Set: How to Achieve Optimality?,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Constrained Policy Optimization via Sampling-Based Weight-Space Projection
SCPO is a sampling-based weight-space projection algorithm that enforces rollout-evaluated safety constraints in policy optimization and provides a safe-by-induction guarantee from any safe starting point.