TextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP

Our summary

A framework that builds input perturbations from a goal function, constraints, a transformation, and a search method, then measures how small, meaning-preserving changes flip a model's output.

Why it matters

Operationalizes perturbation testing — the 'meaning-preserving in, same answer out' relation — at scale, the way to quantify brittleness.

Cited by these methods

Cite this