AI Paper Atlas

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

Identifying Robust Neural Pathways: Few-Shot Adversarial Mask Tuning for Vision-Language Models

Wonjeong Choi, Sejong Ryu, Jungmoon Lee, Dong-Jun Han, Jaekyun Moon

ICLR 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Vision-language model robustness

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 AdvMask mask tuning instead of FAP prompt tuningCLIP ViT-B/32 · base-to-new, new classesadversarial accuracy: 21.9 -> 26.1Improved
use AdvMask instead of FAPCLIP ViT-B/32 · base-to-new, base classesadversarial accuracy: 38.0 -> 43.6Improved
use AdvMask instead of FAP in zero-shot robustness (16-shot)CLIP ViT-B/32 · TinyImageNet to unseen datasetsadversarial accuracy: 16.8 -> 19.4Improved
use AdvMask 16-shot vs TGA-ZSR 16-shotCLIP ViT-B/32 · zero-shot robustnessadversarial accuracy: 13.0 -> 19.4Improved
apply masks to MHSA only instead of MLP only (ablation)AdvMask · 16-shot, 5 datasetsadversarial accuracy: 45.95 -> 47.13Improved
add MLP masks to MHSA masks (ablation)AdvMask · 16-shot, 5 datasetsclean accuracy: 67.34 -> 66.01Hurt
use LAFA loss in addition to adversarial CE (ablation)AdvMask · 16-shot, 5 datasetsadversarial accuracy: 46.4 -> 47.1Improved
add KL divergence auxiliary loss to adversarial CE (ablation)AdvMask · 16-shot, 5 datasetsadversarial accuracy: 46.4 -> 43.6Hurt
add JS divergence auxiliary loss to adversarial CE (ablation)AdvMask · 16-shot, 5 datasetsadversarial accuracy: 46.4 -> 46.5No clear effect
add adaptive weighting to LAFA (ablation)AdvMask · 16-shot, 5 datasetsadversarial accuracy: 46.8 -> 47.1Improved
add adaptive weighting to LAFA in 1-shot (ablation)AdvMask · 1-shot, 5 datasetsadversarial accuracy: 17.8 -> 18.4Improved
combine AdvMask with CoOp prompts, then further train promptsAdvMask + CoOp · 16-shot, 5 datasetsadversarial accuracy: 37.7 -> 44.7Improved

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