一作ACL 2026Poster· Natural Language Processing
Writing- RL : Advancing Long-form Writing via Adaptive Curriculum Reinforcement Learning
Xuanyu Lei · Chenliang Li · Yuning Wu · Kaiming Liu · Weizhou Shen · Peng Li · Ming Yan · Fei Huang · Ya-Qin Zhang · Yang Liu
摘要
Recent advances in Large Language Models (LLMs) have enabled strong performance in long-form writing, but current training paradigms remain limited: Supervised Fine-Tuning (SFT) remains constrained by data saturation and performance ceilings, while Reinforcement Learning with Verifiable Reward (R...