Test fixture
Arithmetic, symbolic steps, and structured problem solving.
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, evidence, constraints. Spec says idempotency key is per order. Trace shows two distinct orders share key abc, while retries for one order use different keys. Identify both violations.
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"per order"
]
},
{
"contains": [
"two distinct orders"
]
},
{
"contains": [
"different keys"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"abc"
]
},
{
"contains": [
"retries"
]
},
{
"contains": [
"violations"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"multi-artifact"
]
},
{
"contains": [
"idempotency"
]
},
{
"contains": [
"evidence"
]
}
]
}
}
}temperature
0
max_tokens
500
timeout (s)
120
type
scored
file
reasoning-mas-idempotency-key.json