FA-100431 / Feature preprocessing fit scope / Open access

Learned missingness indicators uses held-out rows while fitting · case 01

Learned missingness indicators uses held-out rows while fitting.

Verified by executionVariant 1 · 7 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

Fitting reads values outside the designated training rows.

THE FAILURE

Fitting reads values outside the designated training rows.

Unsuccessful approach: Filtering rows after computing the statistic still leaves held-out influence.

Case contract

Train and rows are nonempty rectangular rows. Fit the output feature list as columns containing at least one None in training; emit a 0/1 missing flag for each such column in the same order.

Why this case matters

A small deterministic preprocessing contract makes the fit boundary and transform behavior directly inspectable.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(train, rows):
    cols=[j for j in range(len(train[0])) if any(r[j] is None for r in train) or any(r[j] is None for r in rows)]; return [[int(r[j] is None) for j in cols] for r in rows]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('trained missing column',solve([[0,None],[1,2]],[[3,None]]),[[1]])
check('heldout cannot add column',solve([[0,1],[2,3]],[[None,4]]),[[]])
check('heldout cannot remove column',solve([[0,None]],[[4,5]]),[[0]])
check('multiple learned columns',solve([[None,2,None]],[[1,None,3]]),[[0,0]])
check('empty transform',solve([[None]],[]),[])
check('no learned indicators',solve([[1,2]],[[None,3]]),[[]])
check('singleton fit',solve([[None]],[[None]]),[[1]])
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
trained missing column	`[[1]]`	`[[1]]`	Passed
heldout cannot add column	`[[1]]`	`[[]]`	Failed
heldout cannot remove column	`[[0]]`	`[[0]]`	Passed
multiple learned columns	`[[0, 1, 0]]`	`[[0, 0]]`	Failed
empty transform	`[]`	`[]`	Passed
no learned indicators	`[[1]]`	`[[]]`	Failed
singleton fit	`[[1]]`	`[[1]]`	Passed

SHA-256 / 44f9046f164ddb08c439b7732e9cc739dde6135bb3b27f4153f77b858a372a58

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(train, rows):
    cols=list(range(len(train[0]))); return [[int(r[j] is None) for j in cols] for r in rows]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('trained missing column',solve([[0,None],[1,2]],[[3,None]]),[[1]])
check('heldout cannot add column',solve([[0,1],[2,3]],[[None,4]]),[[]])
check('heldout cannot remove column',solve([[0,None]],[[4,5]]),[[0]])
check('multiple learned columns',solve([[None,2,None]],[[1,None,3]]),[[0,0]])
check('empty transform',solve([[None]],[]),[])
check('no learned indicators',solve([[1,2]],[[None,3]]),[[]])
check('singleton fit',solve([[None]],[[None]]),[[1]])
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
trained missing column	`[[0, 1]]`	`[[1]]`	Failed
heldout cannot add column	`[[1, 0]]`	`[[]]`	Failed
heldout cannot remove column	`[[0, 0]]`	`[[0]]`	Failed
multiple learned columns	`[[0, 1, 0]]`	`[[0, 0]]`	Failed
empty transform	`[]`	`[]`	Passed
no learned indicators	`[[1, 0]]`	`[[]]`	Failed
singleton fit	`[[1]]`	`[[1]]`	Passed

SHA-256 / 8fa1c1fda80ac767100376c996c18d2811920dfdf08586e478b6b2765aa9c3fa

HELD IN THE MEMBER ARCHIVE

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

This mechanism has 7 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

Toy learned missingness indicators; it models column selection state, not the original feature matrix. 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 2026-09-30T16:48:35.363010+00:00.

Case digest / 1543b32fab9f9a0160dfbb1a0fe70d81c610529dc3a9eba79345bdc167ca8b5b