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 module creates a large array and captures it in a closure. The closure is returned as a callback and stored in a global registry. The array is never released. Memory usage grows over time. What is the root cause of the memory leak?
{
"metrics": {
"accuracy": {
"checks": [
{
"contains": [
"closure"
]
},
{
"contains": [
"memory leak"
]
},
{
"contains": [
"global registry"
]
}
]
},
"evidence": {
"checks": [
{
"contains": [
"large array"
]
},
{
"contains": [
"never released"
]
},
{
"contains": [
"captured"
]
}
]
},
"constraint": {
"checks": [
{
"contains": [
"garbage collection"
]
},
{
"contains": [
"strong reference"
]
},
{
"contains": [
"deregister"
]
}
]
}
}
}temperature
0
max_tokens
500
timeout (s)
120
type
scored
file
reason-rc-memory-leak-closure.json