AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

On the Adversarial Robustness of Mixture of Experts

Joan Puigcerver, Rodolphe Jenatton, Carlos Riquelme Ruiz, Pranjal Awasthi, Srinadh Bhojanapalli

NeurIPS 2022 · Adversarial robustness and security › Certified and empirical robustness · approach: Provable robust learning guarantees · 4 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
Replace dense ViT-B/32 with sparse V-MoE-B/32ViT-B/32 · ImageNet adversarial attacksclassification errorImproved
Replace dense ViT-B/32 with sparse V-MoE-B/32ViT-B/32 · JFT-300M adversarial attacksfalse discovery rateImproved
PGD adversarial fine-tuningViT-B/32 and V-MoE-B/32 · ImageNetclassification error: 51.7% and 49.8%Hurt
Increase the total experts from 2 to 16 (ablation)V-MoE-B/32 · JFT-300M adversarial attacksfalse discovery rateImproved
Increase the total experts from 16 to 32 (ablation)V-MoE-B/32 · JFT-300M adversarial attacksfalse discovery rateNo clear effect
Use more than 2 total experts (ablation)V-MoE-B/32 · ImageNet adversarial attacksclassification errorNo clear effect
Train the V-MoE model with PGD adversarial fine-tuning (ablation)V-MoE-B/32 · ImageNet adversarial attacksrouting changesImproved
Include the router auxiliary losses in the attack objective (ablation)V-MoE-B/32 · JFT-300M adversarial attacksfalse discovery rateNo clear effect
Use AutoPGD instead of PGDViT-B/32 · JFT-300M at ℓ∞ = 10⁻³false discovery rate: 0.744 -> 0.757Hurt
Use AutoPGD instead of PGDV-MoE-B/32 · ImageNet adversarial attacksrouting changes: near 0.2 -> around 0.5Hurt

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