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 session starts 10:00, expires after 45 minutes idle, refreshes on activity. Activity at 10:20 and 10:50. Is it valid at 11:30?
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"valid"
]
},
{
"contains": [
"11:35"
]
},
{
"contains": [
"10:50"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"45 minutes"
]
},
{
"contains": [
"refreshes"
]
},
{
"contains": [
"activity"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"stateful"
]
},
{
"contains": [
"expiry"
]
},
{
"contains": [
"idle"
]
}
]
}
}
}temperature
0
max_tokens
500
timeout (s)
120
type
scored
file
reason-state-session-expiry.json