paperbot · PL 论文追踪

全部论文 待读 收藏

P2 DeepState

DeepState
ICSE Proceedings of the 44th International Conference on Software Engineering2022匹配度 2/10 · 论文聚焦 RNN 驱动软件系统的测试覆盖准则与标注数据选择，属于 DNN 测试方向，而非用户关注的静态分析、LLM/Agent 改进静态分析或形式化验证、浮点程序分析、贝叶斯程序分析。虽然标签涉及机器学习与程序语言交叉，但核心问题与画像匹配度低。引用 19 inbox
Zixi Liu, Yang Feng, Yining Yin, Zhenyu Chen
测试、调试与修复机器学习与程序语言
crossref: 2026-09-12

AI 速览

一句话：提出 DeepState，面向 RNN 驱动的软件系统，缓解现有 CNN 覆盖准则难以迁移与标注数据成本高的问题。

深度神经网络在各类软件中表现突出，但由其驱动的系统也可能产生错误行为，带来事故与损失。DNN 驱动软件的测试与优化通常依赖大量标注数据，人工成本高、效率低。已有许多覆盖准则被用于卷积神经网络的数据选择，然而由于循环神经网络的工作机理不同，这些准则难以直接应用。为此，论文提出 DeepState，目标是为 RNN 模型提供测试与优化支持，缓解标注数据依赖并提升测试效率。摘要未展开 DeepState 的具体覆盖定义、算法实现与实验评估，因此其技术细节和有效性尚需结合全文确认。

问题
DNN 驱动软件的测试与优化依赖大量标注数据，成本高、效率低；现有覆盖准则主要面向 CNN，因 RNN 工作机理不同而难以直接适用。
方法
论文提出 DeepState，摘要未详述具体方法；从上下文看，其目标是为 RNN 驱动的软件测试与数据选择提供覆盖准则或支持。
评估
摘要未给出实验设置、对比基线或结果，评估方式无法从摘要确定。
原文摘要（Abstract）

Deep Neural Networks (DNN) have achieved tremendous success in various software applications. However, accompanied by outstanding effectiveness, DNN-driven software systems could also exhibit incorrect behaviors and result in some critical accidents and losses. The testing and optimization of DNN-driven software systems rely on a large number of labeled data that often require many human efforts, resulting in high test costs and low efficiency. Although plenty of coverage-based criteria have been proposed to assist in the data selection of convolutional neural networks, it is difficult to apply them on Recurrent Neural Network (RNN) models due to the difference between the working nature.

链接与引用

DOI 原文 ·

BibTeX 
@inproceedings{paperbot6941,
  title = {DeepState},
  author = {Zixi Liu and Yang Feng and Yining Yin and Zhenyu Chen},
  booktitle = {ICSE},
  year = {2022},
  doi = {10.1145/3510003.3510231}
}