AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

AdvDreamer Unveils: Are Vision-Language Models Truly Ready for Real-World 3D Variations?

Shouwei Ruan, Hanqing Liu, Yao Huang, Xiaoqi Wang, Caixin Kang, Hang Su, Yinpeng Dong, Xingxing Wei

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 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 naturalness reward model (NRM) to adversarial optimization (ablation)AdvDreamer · ImageNet, OpenCLIP ViT-B/16Score_R realism: 1.60 -> 2.52Improved
add naturalness reward model (NRM) (ablation)AdvDreamer · ImageNet, OpenCLIP ViT-B/16Score_P plausibility: 1.39 -> 2.57Improved
add naturalness reward model (NRM) (ablation)AdvDreamer · ImageNet, OpenCLIP ViT-B/16top-1 accuracy (attack strength): 48.6 -> 54.0Hurt
use ISP loss instead of MF_it (ablation)AdvDreamer · ImageNet p*(Theta), OpenCLIP ViT-B/16top-1 accuracy: 58.1% -> 54.0%Improved
use ISP loss instead of MF_ii (ablation)AdvDreamer · ImageNet Theta, OpenCLIP ViT-B/16top-1 accuracy: 22.2% -> 17.4%Improved
use ISP loss instead of MF_ii+MF_tt (ablation)AdvDreamer · ImageNet p*(Theta), OpenCLIP ViT-B/16top-1 accuracy: 54.9% -> 54.0%No clear effect
optimize adversarial distribution p*(Theta) vs random 3D transformOpenCLIP 149M · ImageNet zero-shottop-1 accuracy: 62.6 -> 54.0Improved
use distribution center Theta* instead of sampling p*(Theta)OpenCLIP 149M · ImageNet zero-shottop-1 accuracy: 54.0 -> 18.0Improved
scale up variance of p*(Theta) beyond the optimum (ablation)AdvDreamer · ImageNet, 5 images x 100 samplesattack success rateHurt
increase optimization iterations (ablation)AdvDreamer · ImageNet, OpenCLIP/BLIP ViT-B/16attack success rateImproved
transfer attack with OpenCLIP surrogate to OpenAI CLIP using Theta*AdvDreamer · ImageNet cross-modeltop-1 accuracy: 85.3 -> 29.7Improved
reproduce Adv-3DT samples physicallyAdvDreamer · Physical zero-shot classificationaccuracy: 100.0 -> 51.3Hurt

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