AI Paper Atlas

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

Chain-of-Thought Driven Adversarial Scenario Extrapolation for Robust Language Models

Md Rafi Ur Rashid, Vishnu Asutosh Dasu, Ye Wang, Gang Tan, Shagufta Mehnaz

AAAI 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Adversarial prompts for language models · 3 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
add ASE three-step scenario reasoningGPT-4o · JailBreakV-28kunsafe response rate: 6.25% -> 0.68%Improved
add ASELlama3.3-70B · JailBreakV-28kunsafe response rate: 62.43% -> 3.15%Improved
add ASEGPT-4o · JailBreakV-28koutright rejection: 88.27% -> 10.89%Improved
add ASEGPT-4o · TruthfulQA (adversarial)correctness: 74.37% -> 92.45%Improved
add ASEGPT-4o · RealToxicityPromptstoxicity score: 35.2 -> 5.5Improved
add ASEGPT-4o · HolisticBiasaverage bias score: 27.5 -> 3.3Improved
add ASEGPT-4o · MMLUaccuracy: 78.18% -> 82.04%Improved
add ASEGPT-4o · CNN/DailyMailROUGE-L: 25.67 -> 25.28No clear effect
use Intention Analysis instead of ASEGPT-4o · JailBreakV-28koutright rejection: 97.44% vs ASE 10.89%Hurt
add Paraphrase defenseGPT-4o · JailBreakV-28kunsafe response rate: 6.25% -> 7.43%Hurt
add Paraphrase defenseGPT-4o · TruthfulQA (adversarial)correctness: 74.37% -> 67.73%Hurt
apply ASE to a small modelMistral-7B · JailBreakV-28kunsafe response rate: 61.79% -> 9.87%Mixed

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