Hallucination and unsupported factual assertions · Version 0.1.0
Citation grounding
For answers that cite sources: whether each cited source actually supports the statement it is attached to. Catches confident answers with decorative or mismatched citations.
Execution mode: Synthetic systems only — no live customer endpoints
What it needs to run
- Answer set with citations
- Cited source documents (synthetic corpus)
Permissions and data rules
- Written authorization from the system owner
- Synthetic data and synthetic canaries only — no real personal data
Known limitations
Judge verdicts on partial support are noisy; borderline cases go to human adjudication per methodology.
False-positive risks
Sources that support the gist but not the exact wording can flip the verdict either way.
What is measured
supported-citation rate (judge estimate) — reported per case against the denominator "citations checked". Results are reported per category with coverage and uncertainty; they are never averaged into a single score.
How results are judged
thresholds set per engagement before the run; inconclusive reported separately
Framework mappings
- OWASP LLM Top 10 — LLM09 Misinformation
- NIST AI RMF — MEASURE
- NIST AI RMF — MANAGE
References
Methodology reviewed: 2026-11-08 · Owner: ComplianceEU methodology