ROSE: A Reward-Oriented Data Selection Framework for LLM Task-Specific Instruction Tuning

Open in new window