# FLR012-LAB1 — Claim Grounding & Hallucination Triage

Implement an evidence-bounded `evaluate(case)` function over eight synthetic claim/evidence cases. Treat each claim independently. Your decision must use source authority, evidence entailment, freshness, contradiction evidence, and availability of evidence.

Do not treat citation presence or `model_confidence` as proof. Preserve `UNKNOWN` when evidence is stale, ambiguous, unavailable, or has unresolved authority. The changed-evidence cases must be re-evaluated from their post-evidence fields rather than from the original label. No network, production data, or real high-stakes decision is required.