OpenAI releases MentalHealthBench: 1215 dialogues evaluating AI mental-health response capability
MentalHealthBench was developed with participation of over 80 licensed psychologists, contains 5262 expert-rated criteria, covering 19 languages. GPT-6 Astra scored highest at 57.3%, but reference answers written by clinicians only scored 38.5%, the gap prompts deep thought.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open the original article on ITHome