Exploring Safety Alignment Evaluation of LLMs in Chinese Mental Health Dialogues via LLM-as-Judge

TL;DR

This paper introduces PsyCrisis-Bench, a reference-free evaluation benchmark for assessing the safety alignment of Chinese mental health dialogue models using an LLM-as-Judge approach grounded in psychological principles.

Contribution

It proposes a novel prompt-based LLM-as-Judge method for safety evaluation without gold standards, along with a high-quality Chinese dataset for mental health dialogues.

Findings

Achieves high agreement with expert assessments

Provides interpretable safety evaluation rationales

Outperforms existing evaluation approaches

Abstract

Evaluating the safety alignment of LLM responses in high-risk mental health dialogues is particularly difficult due to missing gold-standard answers and the ethically sensitive nature of these interactions. To address this challenge, we propose PsyCrisis-Bench, a reference-free evaluation benchmark based on real-world Chinese mental health dialogues. It evaluates whether the model responses align with the safety principles defined by experts. Specifically designed for settings without standard references, our method adopts a prompt-based LLM-as-Judge approach that conducts in-context evaluation using expert-defined reasoning chains grounded in psychological intervention principles. We employ binary point-wise scoring across multiple safety dimensions to enhance the explainability and traceability of the evaluation. Additionally, we present a manually curated, high-quality Chinese-language dataset covering self-harm, suicidal ideation, and existential distress, derived from real-world online discourse. Experiments on 3600 judgments show that our method achieves the highest agreement with expert assessments and produces more interpretable evaluation rationales compared to existing approaches. Our dataset and evaluation tool are publicly available to facilitate further research.