AI RESEARCH
Enhancing Operational Safety via Agentic Dialogue Hazard Identification Analysis
arXiv CS.AI
•
ArXi:2606.03812v1 Announce Type: new Operational safety in high-stakes domains such as industrial process control, autonomous, and safety-critical systems, demand reliable hazard identification. While large language models (LLMs) have shown promise in automating safety analysis tasks, single-turn, monolithic inference is brittle: it lacks the self-correction, deliberation, and contextual refinement that safety engineers apply iteratively. In this paper, we