FA-98001 / Training label alignment / Open access

Batch sampler selection is ignored when labels are paired · case 01

Batch sampler selection is ignored when labels are paired.

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

ROOT CAUSE

The pairing routine receives a selected-index list but emits rows from the full batch.

THE FAILURE

The pairing routine receives a selected-index list but emits rows from the full batch.

Unsuccessful approach: Sorting selected indices repairs membership but changes the sampler-requested order.

Case contract

Given examples and labels of equal length plus selected indices, return paired rows in index order.

Why this case matters

A small deterministic training-data operation makes label ownership and ordering explicit.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(examples, labels, indices):
    return [[examples[i],labels[i]] for i in range(len(examples))]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('nonprefix subset',solve(list(range(N+2)),[10+i for i in range(N+2)],[N+1,0]),[[N+1,N+11],[0,10]])
check('reverse selection',solve(['a','b'],[1,2],[1,0]),[['b',2],['a',1]])
check('empty selection',solve(['x'],[N],[]),[])
check('empty batch',solve([],[],[]),[])
check('single selected',solve(['x','y'],[0,N],[1]),[['y',N]])
check('repeated index',solve(['x'],[N],[0,0]),[['x',N],['x',N]])
check('full ordered selection',solve([1,2,3],[4,5,6],[0,1,2]),[[1,4],[2,5],[3,6]])
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
nonprefix subset	`[[0, 10], [1, 11], [2, 12]]`	`[[2, 12], [0, 10]]`	Failed
reverse selection	`[['a', 1], ['b', 2]]`	`[['b', 2], ['a', 1]]`	Failed
empty selection	`[['x', 1]]`	`[]`	Failed
empty batch	`[]`	`[]`	Passed
single selected	`[['x', 0], ['y', 1]]`	`[['y', 1]]`	Failed
repeated index	`[['x', 1]]`	`[['x', 1], ['x', 1]]`	Failed
full ordered selection	`[[1, 4], [2, 5], [3, 6]]`	`[[1, 4], [2, 5], [3, 6]]`	Passed

SHA-256 / 7ef086ae2b79a8c3d8bc9d0ea497d97526174edb2c6ae4c4c7ba372790a3199e

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(examples, labels, indices):
    return [[examples[i],labels[i]] for i in sorted(indices)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('nonprefix subset',solve(list(range(N+2)),[10+i for i in range(N+2)],[N+1,0]),[[N+1,N+11],[0,10]])
check('reverse selection',solve(['a','b'],[1,2],[1,0]),[['b',2],['a',1]])
check('empty selection',solve(['x'],[N],[]),[])
check('empty batch',solve([],[],[]),[])
check('single selected',solve(['x','y'],[0,N],[1]),[['y',N]])
check('repeated index',solve(['x'],[N],[0,0]),[['x',N],['x',N]])
check('full ordered selection',solve([1,2,3],[4,5,6],[0,1,2]),[[1,4],[2,5],[3,6]])
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
nonprefix subset	`[[0, 10], [2, 12]]`	`[[2, 12], [0, 10]]`	Failed
reverse selection	`[['a', 1], ['b', 2]]`	`[['b', 2], ['a', 1]]`	Failed
empty selection	`[]`	`[]`	Passed
empty batch	`[]`	`[]`	Passed
single selected	`[['y', 1]]`	`[['y', 1]]`	Passed
repeated index	`[['x', 1], ['x', 1]]`	`[['x', 1], ['x', 1]]`	Passed
full ordered selection	`[[1, 4], [2, 5], [3, 6]]`	`[[1, 4], [2, 5], [3, 6]]`	Passed

SHA-256 / e0cf9325e2f261c728841c99d01f3e2308b207b8029db6f065f2c83b13c9b12b

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

Indices are valid positions; examples and labels have equal length. 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 2026-09-30T16:48:08.406732+00:00.

Case digest / fb275633fff10849c6afb94ee597dcc840e9251b16d48fcd974d660eeb6b78b0