LLM Judges Fail to Detect Omissions in Clinical Notes

Current LLM judges show significant blindness to omitted information in clinical note audits, often performing near chance on detecting absent facts. The study identifies a critical gap in using second-pass models to verify ambient AI scribes.