FA-97871 / Dataset split isolation / Open access

Dataset split: rolling validation windows reuse future observations as training · case 01

rolling validation windows reuse future observations as training.

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

ROOT CAUSE

Each fold cutoff is computed, but records after its origin are retained.

THE FAILURE

Each fold cutoff is computed, but records after its origin are retained.

Unsuccessful approach: Filtering once against the final origin does not enforce earlier folds.

Case contract

Return a list of sorted training ID lists, one per origin, using records [id,time].

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(records, origins):
    return [sorted(i for i,t in records if t<max(origins)) for _ in origins]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('two origins',solve([['a',N-2],['b',N],['c',N+2]],[N,N+2]),[['a'],['a','b']])
check('empty origins',solve([['a',N]],[]),[])
check('empty data',solve([],[N]),[[]])
check('exact origin excluded',solve([['a',N]], [N]),[[]])
check('unordered records',solve([['c',N+3],['a',N-1],['b',N+1]],[N,N+2]),[['a'],['a','b']])
check('repeated origins',solve([['a',N-1]],[N,N]),[['a'],['a']])
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
two origins	`[['a', 'b'], ['a', 'b']]`	`[['a'], ['a', 'b']]`	Failed
empty origins	`[]`	`[]`	Passed
empty data	`[[]]`	`[[]]`	Passed
exact origin excluded	`[[]]`	`[[]]`	Passed
unordered records	`[['a', 'b'], ['a', 'b']]`	`[['a'], ['a', 'b']]`	Failed
repeated origins	`[['a'], ['a']]`	`[['a'], ['a']]`	Passed

SHA-256 / 81c6f2597cd9820fe895a5f1ba4e21a164418d3b3e13923462034645ae6353ca

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(records, origins):
    return [sorted(i for i,t in records if t<=o) for o in origins]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('two origins',solve([['a',N-2],['b',N],['c',N+2]],[N,N+2]),[['a'],['a','b']])
check('empty origins',solve([['a',N]],[]),[])
check('empty data',solve([],[N]),[[]])
check('exact origin excluded',solve([['a',N]], [N]),[[]])
check('unordered records',solve([['c',N+3],['a',N-1],['b',N+1]],[N,N+2]),[['a'],['a','b']])
check('repeated origins',solve([['a',N-1]],[N,N]),[['a'],['a']])
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
two origins	`[['a', 'b'], ['a', 'b', 'c']]`	`[['a'], ['a', 'b']]`	Failed
empty origins	`[]`	`[]`	Passed
empty data	`[[]]`	`[[]]`	Passed
exact origin excluded	`[['a']]`	`[[]]`	Failed
unordered records	`[['a'], ['a', 'b']]`	`[['a'], ['a', 'b']]`	Passed
repeated origins	`[['a'], ['a']]`	`[['a'], ['a']]`	Passed

SHA-256 / eb27d8953e350bdab8689ef5f827e91da7f44b959322d665b39c0654caf9b4a6

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 2026-09-30T16:48:06.687660+00:00.

Case digest / 5e062f22755ca8d978392c0b88f0a8063268c60c7d240c207b80e8e34a668023