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 numeric parser uses Number(value || 0), silently converting empty strings and null to zero when they should be invalid. Identify the bug and fix.
{
"metrics": {
"repro": {
"checks": [
{
"contains": [
"empty string"
]
},
{
"contains": [
"null"
]
},
{
"contains": [
"zero"
]
},
{
"contains": [
"invalid"
]
}
]
},
"hidden": {
"checks": [
{
"contains": [
"validate"
]
},
{
"contains": [
"Number.isFinite"
]
},
{
"contains": [
"reject empty"
]
}
]
},
"diagnose": {
"checks": [
{
"contains": [
"Number"
]
},
{
"contains": [
"fallback"
]
},
{
"contains": [
"parser"
]
}
]
}
}
}temperature
0
max_tokens
420
timeout (s)
120
type
scored
file
debug-number-parse-empty-v2.json