作者位次 #5/13ACL 2026Poster· Natural Language Processing
SPARD : Self-Paced Curriculum for RL Alignment via Integrating Reward Dynamics and Data Utility
Xuyang Zhi · Peilun Zhou · Chengqiang Lu · Hang Lv · Yiwei Liang · Rongyang Zhang · Yan Gao · Yi Wu · Yao Hu · Hongchao Gu · Defu Lian · Mingjia Yin (University of Science and Technology of China & · Enhong Chen (University of Science and Technology of China &
摘要
The evolution of Large Language Models (LLMs) is shifting the focus from single, verifiable tasks toward complex, open-ended real-world scenarios, imposing significant challenges on the post-training phase. In these settings, the scale and complexity of reward systems have grown significantly, tr...