stability-analysis

该论文研究了分布式训练中外部Nesterov动量在随机伪梯度上的稳定性，推导了随机二次模型上的精确均方稳定性判据，揭示了内层批量大小、本地步数与工作节点平均对训练稳定性的联合影响，并在神经网络上验证了理论结论。

This paper studies stable miscalibration in large language models, where high-confidence errors remain locally stable under perturbations, using diagnostics like audit scores and probes to assess calibration and internal sensitivity.