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. An OpenAPI spec says field status is optional, generated TypeScript marks it required, and logs show clients omitting it. Identify the contract drift and safest fix.
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"optional"
]
},
{
"contains": [
"required"
]
},
{
"contains": [
"contract drift"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"OpenAPI"
]
},
{
"contains": [
"TypeScript"
]
},
{
"contains": [
"clients omitting"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"align generated types"
]
},
{
"contains": [
"backward compatible"
]
},
{
"contains": [
"safest"
]
}
]
}
}
}temperature
0
max_tokens
560
timeout (s)
120
type
scored
file
reasoning-mas-contract-drift.json