Caution for the Environment: Multimodal LLM Agents are Susceptible to Environmental Distractions

This paper highlights the vulnerability of multimodal agents to environmental distractions. The researchers demonstrate that these agents, which process multiple types of input (e.g., text, images, audio), can be significantly impacted by irrelevant or misleading environmental cues. The study provides insights into the limitations of current multimodal systems and emphasizes the need for more robust architectures that can filter out distractions and maintain focus on relevant information in complex, real-world environments.

• MobileJailBench: Red Queen Jailbreak Benchmarking for Mobile Agents
  October 16, 2026 · arXiv
• MMA-SafetyBench: A Benchmark for Multimodal Agent Safety Evaluation
  October 14, 2026 · arXiv
• Control OSWorld: An AI Control Environment for GUI Computer Use Agents
  October 2, 2026 · arXiv
• Learning Reliable GUI Agents under Imperfect Priors
  September 30, 2026 · arXiv
• SCOUT: Synergizing Reasoning and Tool-Use for Computer-Use Safety
  September 28, 2026 · arXiv
• Before Acting, Change the State: Prospective State Intervention for Web Agents under Deceptive Interfaces
  September 28, 2026 · arXiv
• REBASE: Device-Cloud Experience Coherence for GUI Agents Across App Updates
  September 26, 2026 · arXiv
• Constructing Challenging Browser-Use Tasks by Controlled Environment Interventions
  September 20, 2026 · arXiv
• A Dual-Process Perspective on Nudge Susceptibility in LLM-Based GUI Agents
  September 17, 2026 · arXiv
• AgentHijack: Visual Patch Attacks on Multimodal Computer-Use Agents
  September 6, 2026 · arXiv
• WiP: Characterizing and Defending Against Mobile-Agent-Driven MFA Automation
  September 2, 2026 · arXiv