Making a Clinical AI Admit What It Doesn't Know: Grounded Generation With Evidence

Automated excerpt

In a clinical setting, a fluent wrong answer is worse than no answer. A co-pilot for therapists should only recommend what a child's records actually support. Check citations in code: If validation fails, retry once, then fall back to "insufficient records. " Build a small evaluation set of realistic queries, including ones where the records have no answer.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.