FA-97866 / Dataset split isolation / Open access

Dataset split: overlapping windows are retained as training examples · case 01

overlapping windows are retained as training examples.

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

ROOT CAUSE

The overlap predicate identifies intervals that intersect validation and mistakenly returns those conflicts as training data.

THE FAILURE

The overlap predicate identifies intervals that intersect validation and mistakenly returns those conflicts as training data.

Unsuccessful approach: Removing intervals whose start lies inside validation leaves intervals that began earlier but overlap.

Case contract

Intervals [id,start,end] use half-open ranges; keep training intervals with end<=lo or start>=hi.

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(intervals, lo, hi):
    return sorted(i for i,s,e in intervals if s<hi and e>lo)
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('left adjacent',solve([['a',0,N]],N,N+2),['a'])
check('overlap from left',solve([['a',N-2,N+1]],N,N+2),[])
check('inside',solve([['a',N,N+1]],N,N+2),[])
check('right adjacent',solve([['a',N+2,N+4]],N,N+2),['a'])
check('empty',solve([],N,N+2),[])
check('zero length',solve([['a',N,N]],N,N+2),['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
left adjacent	`[]`	`['a']`	Failed
overlap from left	`['a']`	`[]`	Failed
inside	`['a']`	`[]`	Failed
right adjacent	`[]`	`['a']`	Failed
empty	`[]`	`[]`	Passed
zero length	`[]`	`['a']`	Failed

SHA-256 / 14a46e1ec1abeab40d51ac64e4f56af6beb8bbf5fd79d7df26e99dc1c1fa4f71

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(intervals, lo, hi):
    return sorted(i for i,s,e in intervals if not (lo<=s<hi))
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('left adjacent',solve([['a',0,N]],N,N+2),['a'])
check('overlap from left',solve([['a',N-2,N+1]],N,N+2),[])
check('inside',solve([['a',N,N+1]],N,N+2),[])
check('right adjacent',solve([['a',N+2,N+4]],N,N+2),['a'])
check('empty',solve([],N,N+2),[])
check('zero length',solve([['a',N,N]],N,N+2),['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
left adjacent	`['a']`	`['a']`	Passed
overlap from left	`['a']`	`[]`	Failed
inside	`[]`	`[]`	Passed
right adjacent	`['a']`	`['a']`	Passed
empty	`[]`	`[]`	Passed
zero length	`[]`	`['a']`	Failed

SHA-256 / f5b9b374c8929a125b24d394930d8bd9fe1844809103287b697872399b7b6db1

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 2026-09-30T16:48:06.687641+00:00.

Case digest / 3b706de869a6a2f8b99b391dfc9c4f6efdc37a69ccbe5ffeab70b379eeff84f4