Self-play with Execution Feedback: Improving Instruction-following Capabilities of Large Language Models

One core capability of large language models (LLMs) is to follow natural language instructions. However, the issue of automatically constructing high-quality training data to enhance the complex instruction-following abilities of LLMs without manual annotation remains unresolved. In this paper, we introduce AUTOIF, the first scalable and reliable method for automatically generating instructionfollowing training data. AUTOIF transforms the validation of instruction-following data quality into code verification, requiring LLMs to generate instructions, the corresponding code to check the correctness of the instruction responses, and unit test samples to verify the code's correctness. Then, execution feedbackbased rejection sampling can generate data for Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF) training. AUTOIF achieves significant improvements across three training algorithms, SFT, Offline DPO, and Online DPO, when applied to the top open-source LLMs, Qwen2 and LLaMA3, in self-alignment and strong-to-weak distillation settings. Our code is publicly available at https://github.com/QwenLM/AutoIF . Instruction Response Keep your response under 20 characters in length. Are you familiar with OET or Occupational English Test ? Response 1：Yes. Response 2：Yes, I'm familiar with OET. Verification Function Include at least one word ending with '-ing'. What is the weather like today? Response 1：Today's weather is sunny and the wind is blowing. Response 2：The weather is sunny and it is windy today.