General Cooperative Inverse RL Convergence — LessWrong