AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

TrustCLIP: Learning Private Visual Features via Adversarial Reconstruction

Nikos Athanasiou, Ilya A. Petrov, Angela Yao, Shugao Ma, Eric Sauser, Edoardo Remelli, Shreyas Hampali, Johannes Schönberger, Fadime Sener, Bugra Tekin

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 0 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 adversarially trained privacy projection to CLIP featuresTrustCLIP · SUN397DreamSim (privacy): 0.215 -> 0.514Improved
add adversarially trained privacy projection to CLIP featuresTrustCLIP · SUN397Top-1 accuracy: 83.36 -> 82.915No clear effect
use learned projection instead of Gaussian noiseTrustCLIP · SUN397privacy-utility: DSIM 0.68 at 17.4% Top-1 (noise 1.0) vs 0.88 at 85.1%Improved
add Gaussian noise sigma up to 0.1 (ablation)CLIP features · SUN397DreamSim: 0.34 -> 0.33No clear effect
use standard MLP projection instead of identity-initialized (ablation)TrustLLaVA · MM-Vetscore: 34.0 -> 27.8Hurt
use standard MLP projection instead of identity-initialized (ablation)TrustLLaVA · VLM reconstruction attackDreamSim: 0.43 -> 0.62Improved
add TrustCLIP projection to LLaVA-SPTrustLLaVA · MME-Perceptionscore: 1470.7 -> 1390.8Hurt
add TrustCLIP projection to LLaVA-SPTrustLLaVA · LLaVA-Benchscore: 61.7 -> 65.3Improved
add TrustCLIP projection to LLaVA-SPTrustLLaVA · VQAv2score: 79.2 -> 76.3Hurt
add TrustCLIP projection to LLaVA-SPTrustLLaVA · VLM reconstruction attackDreamSim: 0.32 -> 0.39Improved
vary lambda_rec from 0.1 to 0.001 (ablation)TrustLLaVA · VLM benchmarksMM-Vet: 33.1 -> 34.0No clear effect
raise lambda_rec on SUN397 from 0.25 to 1 with L2+LPIPS (ablation)TrustCLIP · SUN397DreamSim: 0.514 -> 0.799Improved

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