FA-100101 / Preference pair construction / Open access

Preference pairs: zero-confidence annotations count as preferences · case 01

zero-confidence annotations count as preferences.

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

ROOT CAUSE

A rating with zero confidence contributes an endpoint despite carrying no preference evidence.

THE FAILURE

A rating with zero confidence contributes an endpoint despite carrying no preference evidence.

Unsuccessful approach: Excluding only negative confidence still includes zero-weight annotations.

Case contract

Rows are [id,score,confidence]. Confidence must be positive; return [best_id,worst_id] among eligible rows, or None below two.

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 [max(rows,key=lambda r:r[1])[0],min(rows,key=lambda r:r[1])[0]] if len(rows)>1 else None
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
a='a'+str(N); b='b'+str(N); x='x'+str(N)
check('zero confidence excluded',solve([[a,N,1],[x,N+5,0],[b,0,1]]),[a,b])
check('negative excluded',solve([[a,N,1],[x,-N,-1],[b,0,2]]),[a,b])
check('one positive',solve([[a,N,1],[b,0,0]]),None)
check('empty',solve([]),None)
check('two positive',solve([[a,N,1],[b,0,1]]),[a,b])
check('all zero',solve([[a,N,0],[b,0,0]]),None)
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
zero confidence excluded	`['x1', 'b1']`	`['a1', 'b1']`	Failed
negative excluded	`['a1', 'x1']`	`['a1', 'b1']`	Failed
one positive	`['a1', 'b1']`	`None`	Failed
empty	`None`	`None`	Passed
two positive	`['a1', 'b1']`	`['a1', 'b1']`	Passed
all zero	`['a1', 'b1']`	`None`	Failed

SHA-256 / 7a5d86beee1fc916804a968c0767e2d6e76063574a3d191a9cbfe5f6f0bd6778

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(rows):
    ok=[r for r in rows if r[2]>=0]
    return [max(ok,key=lambda r:r[1])[0],min(ok,key=lambda r:r[1])[0]] if len(ok)>1 else None
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
a='a'+str(N); b='b'+str(N); x='x'+str(N)
check('zero confidence excluded',solve([[a,N,1],[x,N+5,0],[b,0,1]]),[a,b])
check('negative excluded',solve([[a,N,1],[x,-N,-1],[b,0,2]]),[a,b])
check('one positive',solve([[a,N,1],[b,0,0]]),None)
check('empty',solve([]),None)
check('two positive',solve([[a,N,1],[b,0,1]]),[a,b])
check('all zero',solve([[a,N,0],[b,0,0]]),None)
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
zero confidence excluded	`['x1', 'b1']`	`['a1', 'b1']`	Failed
negative excluded	`['a1', 'b1']`	`['a1', 'b1']`	Passed
one positive	`['a1', 'b1']`	`None`	Failed
empty	`None`	`None`	Passed
two positive	`['a1', 'b1']`	`['a1', 'b1']`	Passed
all zero	`['a1', 'b1']`	`None`	Failed

SHA-256 / 42c87bf712e9895367cab6bd4fed083b8cd6c2b1928b151f3dd33033cf3d6d90

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 2026-09-30T16:48:31.716235+00:00.

Case digest / ce5d483c9d8d159b97eb025c01bfcb98865262d46688bab828fe0f20095a3c27