FA-97896 / Dataset split isolation / Open access

Dataset split: records from one source file are scattered · case 01

records from one source file are scattered.

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

ROOT CAUSE

The split key uses the record ID, not its shared source file.

THE FAILURE

The split key uses the record ID, not its shared source file.

Unsuccessful approach: Grouping by parent directory still combines unrelated files.

Case contract

Rows [id,file]; train files come from the explicit side map; return IDs sorted.

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, sides):
    return sorted(i for i,f in rows if sides.get(i)=='train')
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('same file',solve([['a','dir/a.csv'],['b','dir/a.csv']],{'dir/a.csv':'train'}),['a','b'])
check('different files',solve([['a','a.csv'],['b','b.csv']],{'a.csv':'train','b.csv':'validation'}),['a'])
check('empty',solve([],{}),[])
check('unknown file',solve([['a','x.csv']],{}),[])
check('same directory distinct files',solve([['a','d/a'],['b','d/b']],{'d/a':'train'}),['a'])
check('reversed rows',solve([['b','x'],['a','x']],{'x':'train'}),['a','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
same file	`[]`	`['a', 'b']`	Failed
different files	`[]`	`['a']`	Failed
empty	`[]`	`[]`	Passed
unknown file	`[]`	`[]`	Passed
same directory distinct files	`[]`	`['a']`	Failed
reversed rows	`[]`	`['a', 'b']`	Failed

SHA-256 / aef68df47a1f9abcfdcf185e186bf2a4870b63036b02ec6db4cbb8d8291eec32

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(rows, sides):
    return sorted(i for i,f in rows if sides.get(f.split('/')[0])=='train')
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('same file',solve([['a','dir/a.csv'],['b','dir/a.csv']],{'dir/a.csv':'train'}),['a','b'])
check('different files',solve([['a','a.csv'],['b','b.csv']],{'a.csv':'train','b.csv':'validation'}),['a'])
check('empty',solve([],{}),[])
check('unknown file',solve([['a','x.csv']],{}),[])
check('same directory distinct files',solve([['a','d/a'],['b','d/b']],{'d/a':'train'}),['a'])
check('reversed rows',solve([['b','x'],['a','x']],{'x':'train'}),['a','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
same file	`[]`	`['a', 'b']`	Failed
different files	`['a']`	`['a']`	Passed
empty	`[]`	`[]`	Passed
unknown file	`[]`	`[]`	Passed
same directory distinct files	`[]`	`['a']`	Failed
reversed rows	`['a', 'b']`	`['a', 'b']`	Passed

SHA-256 / 70ab0f14086eb6e57f419ca59ded69c2b856def2a6a64861074c9b350d7afd11

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 2026-09-30T16:48:07.116950+00:00.

Case digest / 1e1dca1547209f18f83beb94b6c76fe5e7a2572a72b48d44b7cba43cd661969e