An open benchmark called MentalHealthBench was introduced, created with the involvement of more than 80 mental health professionals; its goal is to examine the performance of state-of-the-art, so-called frontier models in real conversational situations. The release allows other researchers to verify the methods, run their own evaluations and build further improvements.
AI-generated text
MentalHealthBench released: an open benchmark for mental health dialogues
An open benchmark called MentalHealthBench was introduced, created with the involvement of more than 80 mental health professionals; its goal is to examine the performance of state-of-the-art, so-called frontier models in real conversational situations.
