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. Node stream.pipeline forwards stream errors to its callback or rejected promise. Evaluate a claim that errors are swallowed by default.
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"stream.pipeline"
]
},
{
"contains": [
"errors"
]
},
{
"contains": [
"callback"
]
}
]
},
"fabricated": {
"checks": [
{
"contains": [
"rejected promise"
]
},
{
"contains": [
"not swallowed"
]
},
{
"contains": [
"claim"
]
}
]
}
}
}temperature
0
max_tokens
460
timeout (s)
120
type
scored
file
halluc-api-stream-pipeline.json