Test fixture
Throughput and TTFT-focused generation tasks.
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 benchmark, title, normalized, done. Set benchmark to "speed", title to "Alert Normalization", and done to true. Normalize 18 noisy alert titles into consistent action-oriented summaries.
{
"metrics": {
"complete": {
"checks": [
{
"contains": [
"speed"
]
},
{
"contains": [
"done"
]
},
{
"contains": [
"alert"
]
}
]
}
}
}temperature
0
max_tokens
620
timeout (s)
120
type
throughput
file
speed-alert-normalization.json