Search for a command to run...
Self-play with Execution Feedback: Improving Instruction-following Capabilities of Large Language Models