Test fixture
Grounded answers under adversarial or missing-context prompts.
The model receives the prompt (and optional system message). The run uses scorer rubric_json_metrics with the JSON configuration below. Pass/fail and partial credit are determined entirely by that scorer against the model output; no human grading.
Return JSON only with keys answer, supported_facts, uncertainty. Docs say transform converts payloads to DTO instances when enabled; whitelist strips unknown properties. Assess a claim that whitelist implies transform.
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"transform"
]
},
{
"contains": [
"DTO instances"
]
},
{
"contains": [
"whitelist"
]
}
]
},
"fabricated": {
"checks": [
{
"contains": [
"does not imply"
]
},
{
"contains": [
"unknown properties"
]
},
{
"contains": [
"doc"
]
}
]
}
}
}temperature
0
max_tokens
460
timeout (s)
120
type
scored
file
halluc-doc-validation-pipe.json