Synthesizing Post-Training Data for LLMs through Multi-Agent Simulation

Shuo Tang, Xianghe Pang, Zexi Liu +6

Post-training is essential for enabling large language models (LLMs) to follow human instructions. However, its effectiveness depends on high-quality instruction data, which is cha…