一句话：提出输入反射（Input Reflection），修复导致DNN误预测的失败诱发输入，以缓解有限数据下的分布偏移。
论文关注深度神经网络在真实部署中的可靠性问题：由于训练和测试数据集有限，输入可能偏离训练分布，导致已部署模型产生误预测。标题表明作者提出“输入反射”（Input Reflection）来修复这类失败诱发输入，以缓解有限数据带来的分布外泛化困难。然而，所给摘要仅陈述了问题背景，未说明输入反射的具体机制、算法流程、评估基准或实验结果。因此，本文的核心贡献可概括为面向DNN误预测的失败输入修复方向，但其技术细节与有效性证据需参见全文。
Trained with a sufficiently large training and testing dataset, Deep Neural Networks (DNNs) are expected to generalize. However, inputs may deviate from the training dataset distribution in real deployments. This is a fundamental issue with using a finite dataset, which may lead deployed DNNs to mis-predict in production.
@inproceedings{paperbot7300,
title = {Repairing Failure-inducing Inputs with Input Reflection},
author = {Yan Xiao and Yun Lin and Ivan Beschastnikh and Changsheng Sun and David Rosenblum and Jin Song Dong},
booktitle = {ASE},
year = {2022},
doi = {10.1145/3551349.3556932}
}