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. Two tenants receive webhook events with the same provider event id. Deduplication only keys by event id, and one tenant misses processing. Identify the root cause.
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"tenant"
]
},
{
"contains": [
"event id"
]
},
{
"contains": [
"deduplication"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"same provider"
]
},
{
"contains": [
"misses processing"
]
},
{
"contains": [
"key"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"root cause"
]
},
{
"contains": [
"composite key"
]
},
{
"contains": [
"cross-tenant"
]
}
]
}
}
}temperature
0
max_tokens
500
timeout (s)
120
type
scored
file
reason-root-webhook-dedup.json