Research map / Adversarial robustness and security / Attacks on text and graphs
V-Attack: Targeting Disentangled Value Features for Controllable Adversarial Attacks on LVLMs
Sen Nie, Jie Zhang, Jianxin Yan, Shiguang Shan, Xilin Chen
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 changed | Where | Measured | Effect |
|---|---|---|---|
| target value features instead of patch features (Patch-Attack) (ablation) | V-Attack · ILSVRC2012 CAP avg, LLaVA*, eps=16/255 | attack success rate: 0.388 -> 0.471 | Improved |
| target value features instead of patch features (Patch-Attack) (ablation) | V-Attack · ILSVRC2012 CAP avg, DeepseekVL, eps=16/255 | attack success rate: 0.349 -> 0.421 | Improved |
| use ensemble of surrogates instead of single (ablation) | V-Attack · MS-COCO CAP avg | attack success rate: 0.354 -> 0.567 | Improved |
| use ensemble of surrogates instead of single (ablation) | V-Attack · MS-COCO VQA avg | attack success rate: 0.355 -> 0.560 | Improved |
| use ensemble of surrogates instead of single (ablation) | V-Attack · InternVL CAP (MS-COCO) | attack success rate: 0.283 -> 0.536 | Improved |
| add Value Location (ablation) | V-Attack · MS-COCO VQA, DeepseekVL | attack success rate: 0.463 -> 0.636 | Improved |
| add Self-Value Enhancement (ablation) | V-Attack · MS-COCO VQA, DeepseekVL | attack success rate: 0.521 -> 0.636 | Improved |
| add Self-Value Enhancement (ablation) | V-Attack · MS-COCO CAP, DeepseekVL | attack success rate: 0.560 -> 0.563 | No clear effect |
| increase perturbation budget from 8/255 to 16/255 (ablation) | V-Attack · ILSVRC2012 CAP avg, LLaVA*, V-Attack Ens | attack success rate: 0.352 -> 0.471 | Improved |
| use V-Attack single surrogate vs M-Attack ensemble | V-Attack · MS-COCO CAP, DeepseekVL | attack success rate: 0.483 -> 0.210 | Hurt |
| target value rather than patch features (X-Attack) (ablation) | V-Attack · MS-COCO CAP and VQA | attack success rate | Improved |
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.
- AdvDiff — AdvDiff: Generating Unrestricted Adversarial Examples using Diffusion Models (ECCV 2024) · better on 2 of 2 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
- A Frustratingly Simple Yet Highly Effective Attack Baseline: Over 90% Success Rate Against the Strong Black-box Models of GPT-4.5/4o/o1 (NeurIPS 2025 · similarity 0.91)
- Pandora's Box: Towards Building Universal Attackers against Real-World Large Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- Attention! Your Vision Language Model Could Be Maliciously Manipulated (NeurIPS 2025 · similarity 0.89)
- Chain of Attack: On the Robustness of Vision-Language Models Against Transfer-Based Adversarial Attacks (CVPR 2025 · similarity 0.89)
- Stage-wise Attention-Guided Region Sequencing for Adversarial Attacks on Large Vision-Language Models (NeurIPS 2026 · similarity 0.88)
- SLADE: Shielding against Dual Exploits in Large Vision-Language Models (CVPR 2025 · similarity 0.88)
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.
