ProgramLab

Train single and multi agents to rebuild programs with RL

Blog Code Dataset

1,261 environments where an agent runs a compiled program it cannot read, then rebuilds the whole program. Hidden behavioral tests score the result.