This research establishes a comprehensive framework for evaluating how effectively Large Language Models can detect subtle signs of suicidal thinking and respond appropriately.

• Introduces a novel 1,308-test case dataset built on psychological frameworks
• Evaluates LLMs on two critical capabilities: Identification of Implicit Suicidal ideation (IIS) and Provision of Appropriate Supportive responses (PAS)
• Provides structured assessment methodology for mental health applications of AI
• Offers insights into AI's potential role in suicide prevention technology

This work addresses critical gaps in mental healthcare by exploring how AI could help identify at-risk individuals who may not explicitly express suicidal thoughts, potentially enabling earlier intervention in clinical settings.

Can Large Language Models Identify Implicit Suicidal Ideation? An Empirical Evaluation