When Symptoms Are Not Enough: Evidence-Weighting Patterns in Large Language Model Psychiatric Screening

In three linesSCID-anchored benchmark of 555 semi-structured interviews evaluates 5 LLMs (GPT-4.1 Mini, GPT-5 Mini) on psychiatric screening (anxiety, depression, PTSD). Accuracy 0.49–0.86, MCC 0.16–0.38. False negatives reveal models downweight symptoms when functioning is preserved or social support present, requiring clinical validation before deployment.

Summary generated by Claude — human-verified