Test fixture
Bug fixes, edge conditions, and minimal patch accuracy.
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 diagnosis, fix, tests. A sliding-window rate limiter includes timestamps exactly at the cutoff and returns incorrect allow/deny output at window boundaries. Diagnose and fix.
{
"metrics": {
"repro": {
"checks": [
{
"contains": [
"boundary"
]
},
{
"contains": [
"cutoff"
]
},
{
"contains": [
"timestamp"
]
},
{
"contains": [
"incorrect"
]
}
]
},
"hidden": {
"checks": [
{
"contains": [
">"
]
},
{
"contains": [
"filter"
]
},
{
"contains": [
"window"
]
}
]
},
"diagnose": {
"checks": [
{
"contains": [
"sliding"
]
},
{
"contains": [
"rate limiter"
]
},
{
"contains": [
"off by one"
]
}
]
}
}
}temperature
0
max_tokens
420
timeout (s)
120
type
scored
file
debug-rate-limiter-v2.json