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 counter starts at 0. The following operations are applied in order: add 5, subtract 2, multiply by 3, add 1. What is the final value of the counter?
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"10"
]
},
{
"contains": [
"final value"
]
},
{
"contains": [
"answer"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"subtract 2"
]
},
{
"contains": [
"multiply"
]
},
{
"contains": [
"step"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"order"
]
},
{
"contains": [
"sequential"
]
},
{
"contains": [
"each operation"
]
}
]
}
}
}temperature
0
max_tokens
500
timeout (s)
120
type
scored
file
reason-state-counter-increment.json