Tag
This paper proposes RLAES, a unified LLM framework that jointly optimizes essay scoring and feedback generation via reinforcement learning with rubric-based rewards, achieving state-of-the-art scoring performance on the ASAP benchmark while maintaining high-quality feedback.