PASA-KV: Attacking Action-Facing Visual Routes in World Action Models

Summary

The paper studies adversarial perturbations to the internal visual representations of world action models (WAMs), which couple visual or world representations with action generation. It finds that attack effectiveness varies sharply across visual K/V routes and that larger representation displacement does not reliably imply stronger control failure, attributing this to the action-conditioned readout. The authors introduce Tail-AV, a short-probe signal for identifying vulnerable routes, and PASA-KV, which selects action-facing K/V targets before training a fresh task-shared attack.

Related items

• InfoTextual Response-Aware Guidance for Zero-Shot Adversarial Robustness of Vision-Language ModelsSimilar attack · Arxiv (cs.RO + cs.CV security)
• InfoLiDAR-SHIELD: A Fully Dockerized Adversarial Robustness Assessment Framework for LiDAR-Based 3D Perception ModelsSimilar attack · Arxiv (cs.RO + cs.CV security)
• InfoSynIrisAttack: A benchmark dataset and global-local residual modeling framework for synthetic iris attack detectionSimilar attack · Elsevier Security Journals
• InfoFlipJev: Fast Black-Box Red Teaming Finds Universal Adversarial Suffixes for Typed Decision ModelsSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• InfoAdversarial Self-DistillationSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)