human evaluation

mpathic is a clinician-founded AI safety company that helps AI builders evaluate and improve their models with expert-led evaluation and scientifically grounded benchmarking focused on psychosocial risks. Its platform runs expert judgment across pre-training, post-training and continuously in production, with a network of 7,000+ licensed clinicians writing policy, taxonomy, rubrics and conversation scripts. It offers mPACT benchmarks on suicide risk, eating disorders and misinformation.

Read more

AI-assisted reinforcement learningannotationbenchmarkingexpert-led evaluationmodel red teamingtaxonomy, rubric and policy design