Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Apr 21, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

1.3K

Effectiveness of static text adversarial methods on continuously updating models.

Zihao Yu1, Jun Sun1, Qidong Chen2

  • 1The School of Artificial Intelligence and Computer Science, Jiangnan University, No.1800 Lihu Avenue, Wuxi, 214122, Jiangsu, China.

Neural Networks : the Official Journal of the International Neural Network Society
|April 19, 2026
PubMed
Summary

Related Concept Videos

Survival Tree01:19

Survival Tree

504
Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
 Building a Survival Tree
Constructing a...
504

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Statistics of twinning in strained ferroelastics.

Journal of physics. Condensed matter : an Institute of Physics journal·2017
Same author

Retraction Note to: Relationship between inflammatory cytokines and risk of depression, and effect of depression on the prognosis of high grade glioma patients.

Journal of neuro-oncology·2017
Same author

The Edge Stresses and Phase Transitions for Magnetic BN Zigzag Nanoribbons.

Scientific reports·2017
Same author

Photonics-based broadband radar for high-resolution and real-time inverse synthetic aperture imaging.

Optics express·2017
Same author

Selective malaria antibody screening among eligible blood donors in Jiangsu, China.

Revista do Instituto de Medicina Tropical de Sao Paulo·2017
Same author

Electron Beam Etching of CaO Crystals Observed Atom by Atom.

Nano letters·2017

Static text adversarial attacks struggle against continuously updating natural language processing models. A new evaluation framework reveals significant performance degradation, highlighting the need for dynamic attack strategies.

Area of Science:

  • Natural Language Processing
  • Machine Learning Security

Background:

  • Deep neural networks for natural language processing are vulnerable to adversarial attacks.
  • Existing text adversarial methods are effective against static, fixed models but their performance on dynamic, continuously updated models is unknown.

Purpose of the Study:

  • To investigate the effectiveness of static text adversarial attack methods against continuously updating natural language processing models.
  • To propose a standardized evaluation framework for assessing adversarial attack performance in dynamic model scenarios.

Main Methods:

  • Designed a new task to evaluate static text adversarial attacks on continuously updating models.
  • Developed a comprehensive evaluation framework including novel experimental methods, metrics, and a dynamic baseline.
Keywords:
Black-box attackContinuously updating modelsText adversarial attacking

Related Experiment Videos

Last Updated: Apr 21, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

1.3K
  • Decoupled dynamic attack performance assessment into direct effectiveness and iterative persistence.

    • Main Results:

    • Extensive experiments showed significant performance degradation of static adversarial attack methods against continuously updating models.
    • The proposed framework effectively characterized temporal variations in attack performance.
    • The framework confirmed the necessity of dynamic assessment for adversarial attack evaluation.

    Conclusions:

    • Continuously updating models substantially reduce the effectiveness of static text adversarial attacks.
    • The developed evaluation framework is effective and necessary for precisely characterizing dynamic attack performance in natural language processing.