Test fixture
Cost-aware correctness and efficient API spend per successful task.
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, code, done. Set benchmark to "cost", title to "Implement retry-with-backoff scheduling", and done to true. Produce retry delays for maxAttempts using exponential backoff with cap and optional jitter flag.
{
"metrics": {
"strict": {
"checks": [
{
"contains": [
"cost"
]
},
{
"contains": [
"done"
]
},
{
"contains": [
"backoff"
]
},
{
"contains": [
"cap"
]
},
{
"contains": [
"attempts"
]
}
]
}
}
}temperature
0
max_tokens
520
timeout (s)
150
type
scored
file
cost-complex-retry-with-backoff.json