New Research Targets Security and Long-Horizon Reliability of Vision-Language-Action Models
Two new papers examine VLA models for robotics: one exposes a patch attack exploiting the command-execution gap, while the other introduces progress-aware replanning to improve long-horizon manipulation performance.
Read the full story at arXiv cs.RO ↗
Timeline · 2 reports
- 2026-11-19 05:00 · arXiv cs.RO
RePlan-VLA: Progress-Aware Replanning for Long-Horizon Vision-Language-Action Models - 2026-11-19 05:00 · arXiv cs.RO
ExecGap: Exploiting Command-Execution Gap for Patch Attacks on Vision-Language-Action Models