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. Two concurrent transactions each read a shared counter row, each adds 1 to the value, and each writes back; under Read Committed isolation both read the same initial value so the net increment is 1 instead of 2. Identify the lost-update anomaly and fix it.
{
"metrics": {
"repro": {
"checks": [
{
"contains": [
"concurrent transactions"
]
},
{
"contains": [
"Read Committed"
]
},
{
"contains": [
"counter"
]
},
{
"contains": [
"lost update"
]
}
]
},
"hidden": {
"checks": [
{
"contains": [
"SELECT FOR UPDATE"
]
},
{
"contains": [
"Serializable"
]
},
{
"contains": [
"optimistic locking"
]
}
]
},
"diagnose": {
"checks": [
{
"contains": [
"lost-update anomaly"
]
},
{
"contains": [
"same initial value"
]
},
{
"contains": [
"isolation level"
]
}
]
}
}
}temperature
0
max_tokens
420
timeout (s)
120
type
scored
file
debug-db-transaction-isolation-v2.json