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. A saga reserves inventory, charges payment, then shipment creation fails. Compensation refunds payment but does not release inventory. Trace the remaining inconsistent state.
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"inventory remains reserved"
]
},
{
"contains": [
"refund"
]
},
{
"contains": [
"shipment fails"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"saga"
]
},
{
"contains": [
"compensation"
]
},
{
"contains": [
"state"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"release inventory"
]
},
{
"contains": [
"inconsistent"
]
},
{
"contains": [
"order"
]
}
]
}
}
}temperature
0
max_tokens
560
timeout (s)
120
type
scored
file
reason-state-saga-compensation.json