Shaeke Salman

About

Shaeke Salman is a rising researcher at the forefront of trustworthy AI and multimodal systems, with a primary focus on the security and robustness of vision-language navigation (VLN). In their most-cited work, "Malicious Path Manipulations via Exploitation of Representation Vulnerabilities of Vision-Language Navigation Systems" (2024, 5 citations), Salman exposes critical security flaws in VLN agents that leverage large language models and multi-modal vision-language transformers. They demonstrate how these systems, despite their unprecedented zero-shot recognition and command understanding capabilities, are susceptible to adversarial path manipulations—an essential contribution to the emerging field of AI safety in embodied agents. This work has quickly garnered attention for its practical implications in deploying VLN in real-world environments. Salman’s research sits at the intersection of natural language processing, computer vision, and adversarial machine learning, offering a cautionary yet constructive perspective on the vulnerabilities inherent in cutting-edge multimodal AI. Their findings are vital for researchers and practitioners building next-generation autonomous navigation systems, emphasizing that robustness must keep pace with capability.

Research Focus

Key Achievements

Top Papers

1. 1
   Malicious Path Manipulations via Exploitation of Representation Vulnerabilities of Vision-Language Navigation Systems

   5 citations · 2024

Key Collaborators

Contact & Links