Synthetic Data RL: Task Definition Is All You Need

Yiduo Guo, Zhen Guo, Chuanwei Huang +5

Reinforcement learning (RL) is a powerful way to adapt foundation models to specialized tasks, but its reliance on large-scale human-labeled data limits broad adoption. We introduc…