A hybrid safe-RL framework combining explainable abstraction, formal model checking, and risk-guided search reports more diverse policy failures, but its headline PAC guarantee largely restates its own assumptions.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Verification-Guided Falsification for Safe RL via Explainable Abstraction and Risk-Aware Exploration
A hybrid safe-RL framework combining explainable abstraction, formal model checking, and risk-guided search reports more diverse policy failures, but its headline PAC guarantee largely restates its own assumptions.