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 user has a 10GB disk quota. They have already used 9.8GB. They attempt to upload a 300MB file. Should the upload be allowed or denied?
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"deny"
]
},
{
"contains": [
"exceeds"
]
},
{
"contains": [
"10GB"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"9.8"
]
},
{
"contains": [
"300MB"
]
},
{
"contains": [
"10.1"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"quota"
]
},
{
"contains": [
"exceeds limit"
]
},
{
"contains": [
"over capacity"
]
}
]
}
}
}temperature
0
max_tokens
500
timeout (s)
120
type
scored
file
reason-constraint-disk-quota.json