FA-100116 / Preference pair construction / Open access

Preference pairs: pair generation favors one orientation · case 01

pair generation favors one orientation.

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

ROOT CAUSE

Pair generation keeps only one arbitrary input ordering, biasing the dataset toward one side.

THE FAILURE

Pair generation keeps only one arbitrary input ordering, biasing the dataset toward one side.

Unsuccessful approach: Mirroring text but copying the full original weight doubles total sample mass.

Case contract

Rows are [chosen,rejected,weight]. Return each row and its mirrored orientation, each with half its input weight, preserving row order.

Why this case matters

A small deterministic preference-data transform makes pair semantics inspectable without a training framework.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(rows):
    return [[r[0],r[1],r[2]] for r in rows]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('single mirrored pair',solve([['a','b',N]]),[['a','b',[N,2]],['b','a',[N,2]]])
check('two rows preserve order',solve([['a','b',2],['c','d',4]]),[['a','b',[2,2]],['b','a',[2,2]],['c','d',[4,2]],['d','c',[4,2]]])
check('empty',solve([]),[])
check('weight zero',solve([['a','b',0]]),[['a','b',[0,2]],['b','a',[0,2]]])
check('odd weight exact',solve([['a','b',3]]),[['a','b',[3,2]],['b','a',[3,2]]])
check('same labels',solve([['x','x',N]]),[['x','x',[N,2]],['x','x',[N,2]]])
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
single mirrored pair	`[['a', 'b', 1]]`	`[['a', 'b', [1, 2]], ['b', 'a', [1, 2]]]`	Failed
two rows preserve order	`[['a', 'b', 2], ['c', 'd', 4]]`	`[['a', 'b', [2, 2]], ['b', 'a', [2, 2]], ['c', 'd', [4, 2]], ['d', 'c', [4, 2]]]`	Failed
empty	`[]`	`[]`	Passed
weight zero	`[['a', 'b', 0]]`	`[['a', 'b', [0, 2]], ['b', 'a', [0, 2]]]`	Failed
odd weight exact	`[['a', 'b', 3]]`	`[['a', 'b', [3, 2]], ['b', 'a', [3, 2]]]`	Failed
same labels	`[['x', 'x', 1]]`	`[['x', 'x', [1, 2]], ['x', 'x', [1, 2]]]`	Failed

SHA-256 / 2dbec0c302774b4e1ea729ce1d0ffb54733267314b10e0b5bc218938c89ef9b1

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(rows):
    return [[r[0],r[1],r[2]] for r in rows for _ in (0,1)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('single mirrored pair',solve([['a','b',N]]),[['a','b',[N,2]],['b','a',[N,2]]])
check('two rows preserve order',solve([['a','b',2],['c','d',4]]),[['a','b',[2,2]],['b','a',[2,2]],['c','d',[4,2]],['d','c',[4,2]]])
check('empty',solve([]),[])
check('weight zero',solve([['a','b',0]]),[['a','b',[0,2]],['b','a',[0,2]]])
check('odd weight exact',solve([['a','b',3]]),[['a','b',[3,2]],['b','a',[3,2]]])
check('same labels',solve([['x','x',N]]),[['x','x',[N,2]],['x','x',[N,2]]])
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
single mirrored pair	`[['a', 'b', 1], ['a', 'b', 1]]`	`[['a', 'b', [1, 2]], ['b', 'a', [1, 2]]]`	Failed
two rows preserve order	`[['a', 'b', 2], ['a', 'b', 2], ['c', 'd', 4], ['c', 'd', 4]]`	`[['a', 'b', [2, 2]], ['b', 'a', [2, 2]], ['c', 'd', [4, 2]], ['d', 'c', [4, 2]]]`	Failed
empty	`[]`	`[]`	Passed
weight zero	`[['a', 'b', 0], ['a', 'b', 0]]`	`[['a', 'b', [0, 2]], ['b', 'a', [0, 2]]]`	Failed
odd weight exact	`[['a', 'b', 3], ['a', 'b', 3]]`	`[['a', 'b', [3, 2]], ['b', 'a', [3, 2]]]`	Failed
same labels	`[['x', 'x', 1], ['x', 'x', 1]]`	`[['x', 'x', [1, 2]], ['x', 'x', [1, 2]]]`	Failed

SHA-256 / a77a90c801b6917a4e36fc5539f613b9223a20d8364a92f26f03031d3e36ff55

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

These are explicit toy record contracts; they do not claim compatibility with a trainer or improved model behavior. 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 2026-09-30T16:48:31.747774+00:00.

Case digest / f50da525e257dcb0db3c820f2a60d938d960592c8368516f1913c4dd388465ba