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, rows, done. Set benchmark to "speed", title to "Key Value Table", and done to true. Produce 14 configuration key rows with key, default, and purpose.
{
"metrics": {
"complete": {
"checks": [
{
"contains": [
"speed"
]
},
{
"contains": [
"done"
]
},
{
"contains": [
"key"
]
}
]
}
}
}temperature
0
max_tokens
360
timeout (s)
120
type
throughput
file
speed-key-value-table.json