NeurIPS

Training-Time Explainability for Multilingual Hate Speech Detection: Aligning Model Reasoning with Human Rationales

多语言仇恨言论检测的训练时可解释性：对齐模型推理与人类理由

Muhammad Deedahwar Mazhar Qureshi, Sannaan Khan, Muhammad Atif Qureshi, Wael Rashwan

机构 * Technological University Dublin（都柏林理工大学） ； National University of Sciences and Technology（国家科技大学） ； Maynooth University（梅努斯大学） 纠错

AI总结 该研究针对多语言仇恨言论检测的不透明问题，提出训练时可解释性框架，对齐模型推理与人类理由，经HateXplain、BullySent数据集及多方法评估，可提升分类性能与解释质量，助力多语言文化敏感的内容审核。

Comments Accepted at NeurIPS Workshops 2025

Randomized Quasi-Gauss--Newton Methods for Solving General Nonlinear Equations

求解一般非线性方程的随机拟高斯-牛顿方法

Chengchang Liu, Luo Luo

AI总结 本文提出基于格拉姆矩阵近似的随机拟高斯-牛顿方法，可处理欠定与超定一般非线性方程，欠定情形下局部超线性收敛到最优解，超定情形下无依赖条件数收敛到驻点，优于现有拟牛顿方法。

Comments The conference version was in the proceedings of NeurIPS 2022, where we focus on the special case $n=d$. The results that are generalized to general nonlinear equations were presented at ICCOPT 2025 and ICOTA 2026