To Augment or Not to Augment? Diagnosing Distributional Symmetry Breaking
增强还是不增强?诊断分布对称性破坏
机构 * MIT(麻省理工学院) ; University College London(伦敦大学学院) ; Northeastern University(东北大学)
AI总结 本文提出了一种评估数据集对称性破坏的指标,揭示了点云数据集中的严重偏见问题,并展示了对称性破坏对模型性能的影响。
Comments Published as a conference paper at ICLR 2026. A short version of this paper appeared at the ICLR AI4Mat workshop in April 2025