Question: Reward Hacking Detection Metrics in RL-Trained Agents

Constrained RL penalty coefficient for safe exploration

In audit

0/5 publication-ready

Findings metrics

—

total

—

verified

—

pending

—

rejected