AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Guided Perturbation Sensitivity (GPS): Detecting Adversarial Text via Embedding Stability and Word Importance

Bryan E. Tuck, Rakesh M. Verma

AAAI 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Defenses against universal perturbations · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use gradient attribution instead of random word ranking (ablation)GPS · Adversarial detection (Table 3)detection accuracy: 0.694 -> 0.836 (Yelp RoBERTa BERT-Attack)Improved
use gradient attribution instead of attention rollout (ablation)GPS · Adversarial detection (word-level attacks)detection accuracy: 0.714 -> 0.836 (Yelp RoBERTa BERT-Attack)Improved
use attention rollout instead of gradient on char-level attack (ablation)GPS · IMDB DeepWordBugdetection accuracy: 0.935 vs 0.958 (RoBERTa)No clear effect
guided importance masking vs random masking: sensitivity ratio (ablation)GPS features · Embedding sensitivity (Table 1)adv/benign ratio: 1.932 vs 1.880No clear effect
GPS (Gradient) vs TextShieldGPS · IMDB DeBERTa BERT-Attackdetection accuracy: 0.731 -> 0.755Improved
GPS (Gradient) vs Sharpness baselineGPS · IMDB DeBERTa BERT-Attackdetection accuracy: 0.797 vs 0.755Hurt
GPS (Gradient) vs Sharpness baselineGPS · Yelp RoBERTa BERT-Attackdetection accuracy: 0.910 vs 0.836Hurt
transfer detector from DeepWordBug to BERT-AttackGPS (Gradient) · Cross-attack transferF1: 0.943 -> 0.649Hurt
reduce masked words K from 50 to 5 (ablation)GPS · Detection F1F1: <0.015 F1 variationNo clear effect
train on DeBERTa test on RoBERTaGPS (Gradient) · Cross-model transferF1: 0.852 -> 0.880Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy