ACPO: Adaptive Curriculum Policy Optimization for Aligning Vision-Language Models in Complex Reasoning

Published in arXiv:2510.00690, 2025

小米小爱合作 Xiaomi Collaboration · arXiv 2025

  • 自适应课程策略优化(ACPO),用于复杂推理场景下的视觉–语言模型对齐
  • 与小米 Peng Gao、Bin Qin 等合作
  • 与 GUI Agent / ScreenR1 方向的研究一脉相承

🔗 arXiv

Recommended citation: Y Wang, S Chen, Z Li, T Liu, C Song, J Jiang, J Zhu, P Gao, B Qin. "ACPO: Adaptive Curriculum Policy Optimization for Aligning Vision-Language Models in Complex Reasoning." arXiv:2510.00690, 2025.
Download Paper