FA-100066 / Preference pair construction / Open access

Preference pairs: parallel candidate arrays drift out of alignment · case 01

parallel candidate arrays drift out of alignment.

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

ROOT CAUSE

A positional zip silently truncates candidates when one metadata array is short.

THE FAILURE

A positional zip silently truncates candidates when one metadata array is short.

Unsuccessful approach: Padding the shorter list with its last element fabricates a preference for unrelated records.

Case contract

Chosen and rejected rows are [pair_id,response]. Return [pair_id,chosen,rejected] for shared IDs in chosen input 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(chosen, rejected):
    return [[c[0],c[1],r[1]] for c,r in zip(chosen,rejected)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
p='p'+str(N); q='q'+str(N)
check('reordered keys',solve([[p,'yes'],[q,'up']],[[q,'down'],[p,'no']]),[[p,'yes','no'],[q,'up','down']])
check('missing second match',solve([[p,'yes'],[q,'up']],[[p,'no']]),[[p,'yes','no']])
check('empty both',solve([],[]),[])
check('empty rejected',solve([[p,'yes']],[]),[])
check('empty chosen',solve([],[[p,'no']]),[])
check('unmatched then matched',solve([[q,'up'],[p,'yes']],[[p,'no']]),[[p,'yes','no']])
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
reordered keys	`[['p1', 'yes', 'down'], ['q1', 'up', 'no']]`	`[['p1', 'yes', 'no'], ['q1', 'up', 'down']]`	Failed
missing second match	`[['p1', 'yes', 'no']]`	`[['p1', 'yes', 'no']]`	Passed
empty both	`[]`	`[]`	Passed
empty rejected	`[]`	`[]`	Passed
empty chosen	`[]`	`[]`	Passed
unmatched then matched	`[['q1', 'up', 'no']]`	`[['p1', 'yes', 'no']]`	Failed

SHA-256 / 725b31a10e2356544915c0a12bf52dae5b61743a5e652766ea6d5caf71b2bcdb

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(chosen, rejected):
    rmap={r[0]:r[1] for r in rejected}
    return [[c[0],c[1],rmap.get(c[0],rejected[-1][1] if rejected else None)] for c in chosen]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
p='p'+str(N); q='q'+str(N)
check('reordered keys',solve([[p,'yes'],[q,'up']],[[q,'down'],[p,'no']]),[[p,'yes','no'],[q,'up','down']])
check('missing second match',solve([[p,'yes'],[q,'up']],[[p,'no']]),[[p,'yes','no']])
check('empty both',solve([],[]),[])
check('empty rejected',solve([[p,'yes']],[]),[])
check('empty chosen',solve([],[[p,'no']]),[])
check('unmatched then matched',solve([[q,'up'],[p,'yes']],[[p,'no']]),[[p,'yes','no']])
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
reordered keys	`[['p1', 'yes', 'no'], ['q1', 'up', 'down']]`	`[['p1', 'yes', 'no'], ['q1', 'up', 'down']]`	Passed
missing second match	`[['p1', 'yes', 'no'], ['q1', 'up', 'no']]`	`[['p1', 'yes', 'no']]`	Failed
empty both	`[]`	`[]`	Passed
empty rejected	`[['p1', 'yes', None]]`	`[]`	Failed
empty chosen	`[]`	`[]`	Passed
unmatched then matched	`[['q1', 'up', 'no'], ['p1', 'yes', 'no']]`	`[['p1', 'yes', 'no']]`	Failed

SHA-256 / 39c6217b3fea5b789f0a57bfd7174a74a7d1ce1c5563522f8de467435e419b2f

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 2026-09-30T16:48:31.261688+00:00.

Case digest / 2a1a0f5bab7c6e9e6f2e1dc909a457209713a4a4a58515565127d40161159ea6