Evaluating Large Language Models in Crisis Detection: A Real-World Benchmark from Psychological Support Hotlines

TL;DR

This study evaluates large language models' effectiveness in crisis detection tasks using a real-world benchmark from psychological hotlines, highlighting their potential and limitations in clinical settings.

Contribution

Introduces PsyCrisisBench, a comprehensive benchmark for assessing LLMs in crisis detection, with evaluations across multiple models and tasks in a real-world context.

Findings

LLMs achieved high F1 scores in suicidal ideation detection and risk assessment.

Few-shot prompting and fine-tuning significantly improved LLM performance.

A fine-tuned 1.5B-parameter model outperformed larger models on some tasks.

Abstract

Psychological support hotlines serve as critical lifelines for crisis intervention but encounter significant challenges due to rising demand and limited resources. Large language models (LLMs) offer potential support in crisis assessments, yet their effectiveness in emotionally sensitive, real-world clinical settings remains underexplored. We introduce PsyCrisisBench, a comprehensive benchmark of 540 annotated transcripts from the Hangzhou Psychological Assistance Hotline, assessing four key tasks: mood status recognition, suicidal ideation detection, suicide plan identification, and risk assessment. 64 LLMs across 15 model families (including closed-source such as GPT, Claude, Gemini and open-source such as Llama, Qwen, DeepSeek) were evaluated using zero-shot, few-shot, and fine-tuning paradigms. LLMs showed strong results in suicidal ideation detection (F1=0.880), suicide plan identification (F1=0.779), and risk assessment (F1=0.907), with notable gains from few-shot prompting and fine-tuning. Compared to trained human operators, LLMs achieved comparable or superior performance on suicide plan identification and risk assessment, while humans retained advantages on mood status recognition and suicidal ideation detection. Mood status recognition remained challenging (max F1=0.709), likely due to missing vocal cues and semantic ambiguity. Notably, a fine-tuned 1.5B-parameter model (Qwen2.5-1.5B) outperformed larger models on mood and suicidal ideation tasks. LLMs demonstrate performance broadly comparable to trained human operators in text-based crisis assessment, with complementary strengths across task types. PsyCrisisBench provides a robust, real-world evaluation framework to guide future model development and ethical deployment in clinical mental health.