Test fixture
Grounded answers under adversarial or missing-context prompts.
The model receives the prompt (and optional system message). The run uses scorer exact_match with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.
Answer only based on the document. If not present, reply exactly: Not stated Document: - Product Orion supports SSO via SAML and OIDC. - Backups run every 4 hours and are encrypted at rest. - Audit logs are searchable for 120 days. - API access uses scoped tokens and cursor pagination. - Incident updates are published on the public status page. Question: Which data residency regions are available for tenants?
{
"expected": "Not stated"
}temperature
0
max_tokens
32
timeout (s)
120
type
scored
file
hallucination_hard_01.json