AI Paper Atlas

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

The Latent Guardian: Defending Collaborative Perception via Feature-Level Consistency Verification

Zhuangzhuang Zhang, Mingxin Li, Libing Wu, Wei-Bin Lee, Jianping Wang

ICML 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Attacks and defenses in sequential settings

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 semantic consistency to topological-only check (ablation)Cerberus · MOR (AttFusion)ORR: 8.37% -> 7.94%Improved
add energy consistency to topological-only check (ablation)Cerberus · MOR (AttFusion)ORR: 8.37% -> 7.69%Improved
combine all three consistency metrics (ablation)Cerberus · MOR (AttFusion)ORR: 8.37% -> 7.54%Improved
combine all three consistency metrics (ablation)Cerberus · TOR (AttFusion)ASR: 9.46% -> 8.92%Improved
use adaptive uncertainty-aware threshold instead of fixed (ablation)Cerberus · mixed attack with 50% noisy benign agentsAP@0.5/0.7Improved
increase adaptive attack iterations from 10 to 20Cerberus · MOR (Table 5)ORR: 7.53% -> 8.04%No clear effect
add 5 random restarts to adaptive attackCerberus · MOR (Table 5)ORR: 8.04% -> 8.05%No clear effect
use grouping-aware surrogate with 20 iterations and 5 restartsCerberus · Table 5AP@0.5: 0.84 -> 0.83No clear effect
use Cerberus instead of CP-Guard under TOR (In)CP-Guard · TOR (In) AttFusionASR: 97.42% -> 4.19%Improved
use Cerberus instead of LUCIALUCIA · MOR V2VAMORR: 19.09% -> 7.87%Improved
use Cerberus instead of CP-Guard under MORCP-Guard · MOR AttFusionORR: 7.82% -> 7.54%No clear effect
use Cerberus instead of ROBOSAC under BIMROBOSAC · BIM V2VAMAP@0.5: 0.77 -> 0.88Improved

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