Search for a command to run...
LAPO: Internalizing Reasoning Efficiency via Length-Adaptive Policy Optimization