FA-97891 / Dataset split isolation / Open access

Dataset split: domain holdout leaks through a second domain label · case 01

domain holdout leaks through a second domain label.

Verified by executionVariant 1 · 6 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

Each row can belong to multiple domains and one matching label is missed.

THE FAILURE

Each row can belong to multiple domains and one matching label is missed.

Unsuccessful approach: Using only the first domain misses rows where the held-out domain appears later.

Case contract

Rows [id,domains]; hold out the lexically greatest domain and return rows containing it.

Why this case matters

This toy split allocator makes membership rules explicit for reproducible evaluation fixtures.

1 / The failure

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(rows):
    held=max((d[0] for i,d in rows if d),default='')
    return sorted(i for i,d in rows if d and d[0]==held)
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('later domain member',solve([['a',['x','z']],['b',['z']]]),['a','b'])
check('empty',solve([]),[])
check('empty memberships',solve([['a',[]],['b',[]]]),[])
check('singleton',solve([['a',['z']]]),['a'])
check('overlap domain',solve([['a',['a','z']],['b',['b']]]),['a'])
check('held domain only',solve([['a',['x']],['b',['y','z']]]),['b'])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
later domain member	`['b']`	`['a', 'b']`	Failed
empty	`[]`	`[]`	Passed
empty memberships	`[]`	`[]`	Passed
singleton	`['a']`	`['a']`	Passed
overlap domain	`['b']`	`['a']`	Failed
held domain only	`['b']`	`['b']`	Passed

SHA-256 / 644aceb835296675c4b68478849ca0ba74740063e93994c3c7d68a1b74e26ce4

2 / The unsuccessful fix

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(rows):
    held=max((x for i,d in rows for x in d),default='')
    return sorted(i for i,d in rows if held in d and d[0]==held)
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('later domain member',solve([['a',['x','z']],['b',['z']]]),['a','b'])
check('empty',solve([]),[])
check('empty memberships',solve([['a',[]],['b',[]]]),[])
check('singleton',solve([['a',['z']]]),['a'])
check('overlap domain',solve([['a',['a','z']],['b',['b']]]),['a'])
check('held domain only',solve([['a',['x']],['b',['y','z']]]),['b'])
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)

Boundary fixture	Actual	Expected	Outcome
later domain member	`['b']`	`['a', 'b']`	Failed
empty	`[]`	`[]`	Passed
empty memberships	`[]`	`[]`	Passed
singleton	`['a']`	`['a']`	Passed
overlap domain	`[]`	`['a']`	Failed
held domain only	`[]`	`['b']`	Failed

SHA-256 / 1ccc573658c1316e53c36fb5e6f9a50e64f56910cea3d6ade9fb7bc93e6e1521

HELD IN THE MEMBER ARCHIVE

The verified repair and its recorded checks are member-only.

This mechanism has 6 recorded checks per implementation. The open-access tier publishes the failure and the unsuccessful fix; the repaired source that passes every check, and the observations that prove it, are available to members.

Every case sharing this mechanism uses the same contract and the same repair, so this one record is held back for all of them.

Member access is invitation-based. Sign in with your invited account to inspect the repair.

Verification & scope

Inputs are small in-memory records with self-contained contracts; no claim is made about production sampling quality. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.

Observations recorded using Python 3.12.14 at 2026-09-30T16:48:07.116943+00:00.

Case digest / 7ef6728cb4de1520baf277cc083e15f2358f1f47611cf8951cab6d970b842e17