AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

GraphShield: Graph-Theoretic Modeling of Network-Level Dynamics for Robust Jailbreak Detection

Sunghee Dong, Sungwon Yi, Kangmin Bae, Jaeyoon Kim, Seongyeop Kim

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Jailbreak attacks and defenses

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
concatenate features from all anchors instead of the single anchor can (ablation)GraphShield (LLaMA-2) · JailbreakBench detectionFPR (%): 13.67 -> 7.08Improved
add anchor conditioning to the global graph (ablation)GraphShield (LLaMA-2) · JailbreakBench detectionTPR (%): 68.25 -> 91.00Improved
add graph structure to anchor features instead of pooling (ablation)GraphShield (LLaMA-2) · JailbreakBench detectionTPR (%): 88.82 -> 91.00Improved
add graph modeling to a cosine-only linear probe (ablation)GraphShield (LLaMA-2) · JailbreakBench detectionTPR (%): 83.94 -> 91.00Improved
add anchor tokens to a sequence-level attention graph (ablation)GraphShield (LLaMA-2) · JailbreakBench detectionTPR (%): 87.27 -> 91.00Improved
use default refusal anchors instead of anti-anchors (ablation)GraphShield (LLaMA-2) · JailbreakBench detectionTPR (%): 83.70 -> 91.00Improved
concatenate all feature groups instead of individual groups (ablation)GraphShield · JailbreakBench detectionFPR (%): LLaMA-2: 91.00% / 7.08%; Vicuna: 89.82% / 6.83%Improved
augment detector with a few adaptive attack examplesGraphShield · Adaptive attacksTPR (%): below 6% -> above 90%Improved
evaluate leave-one-attack-out (unseen) instead of seen attacks (ablation)GraphShield · JailbreakBench detectionTPR (%): JOOD TPR < 3%, Decipher about 36-37%Hurt

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