AI Paper Atlas

Research map / Adversarial robustness and security / Black-box and transfer attacks

Dissecting the Safety Circuit: Neuronal Intervention for Transferable Adversarial Attacks on VLMs

Chunlong Xie, Kangjie Chen, Shangwei Guo, Shudong Zhang, Jiamou Liu, Tianwei Zhang, Tao Xiang

ICML 2026 · Adversarial robustness and security › Black-box and transfer attacks · approach: Transfer attacks on face recognition

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
remove defensive circuit suppression loss (ablation)SCIA · Qwen-2.5-VL targetASR: 0.487 -> 0.651Improved
add transferable circuit amplification (ablation)SCIA · Qwen-2.5-VL targetASR: 0.558 -> 0.651Improved
add contrastive semantic steering loss (ablation)SCIA · InstructBlip targetASR: 0.378 -> 0.517Improved
add spectral smoothness regularization (ablation)SCIA · Qwen-2.5-VL targetASR: 0.598 -> 0.651Improved
increase perturbation budget from 4/255 to 16/255 (ablation)SCIA · 3 victim VLMs averageASR: 0.220 -> 0.546Improved
increase perturbation budget from 16/255 to 32/255 (ablation)SCIA · 3 victim VLMs averageASR: 0.546 -> 0.569No clear effect
increase circuit size from 0.5% to 5% of layer neurons (ablation)SCIA · Average over victimsASR: 0.462 -> 0.334Hurt
use very small circuit (0.05% of neurons) (ablation)SCIA · Average over victimsASR: 0.363 vs 0.462 at 0.5%Hurt
deactivate identified defensive circuit neurons (ablation)LLaVA-v1.6 · AdvBenchASR: 0.24 -> 0.86 (random 0.26)Improved
deactivate identified transferable circuit neurons (ablation)surrogate-generated adversarial examples · LLaVA-v1.6 victimASR: 0.79 -> 0.33 (random 0.74)Hurt
apply Gaussian smoothing as input preprocessing defenseSCIA attack · 3 black-box victimsASR: 0.546 -> 0.319Hurt
apply bit-depth reduction as defenseSCIA attack · 3 black-box victimsASR: 0.546 -> 0.473No clear effect

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