FA-97996 / Training label alignment / Open access
Expanded views lose the source example label · case 01
Expanded views lose the source example label.
ROOT CAUSE
Each derived view is emitted without carrying its source supervision value.
THE FAILURE
Each derived view is emitted without carrying its source supervision value.
Unsuccessful approach: Copying the label only to the first view leaves the remaining training rows unlabeled.
Case contract
Examples are [source_id,label,view_ids]. Emit [view_id,label] for every view in source and view 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):
return [[v,None] for e in examples for v in e[2]]
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('all views inherit',solve([['s',N,['v1','v2']]]),[['v1',N],['v2',N]])
check('multiple sources',solve([['a',N,['a1']],['b',N+1,['b1','b2']]]),[['a1',N],['b1',N+1],['b2',N+1]])
check('no views',solve([['x',0,[]]]),[])
check('empty sources',solve([]),[])
check('zero label copied',solve([['x',0,[0,N]]]),[[0,0],[N,0]])
check('single view',solve([['x',N,['only']]]),[['only',N]])
check('source order',solve([['b',2,['b0']],['a',1,['a0']]]),[['b0',2],['a0',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 |
|---|---|---|---|
| all views inherit | [['v1', None], ['v2', None]] | [['v1', 1], ['v2', 1]] | Failed |
| multiple sources | [['a1', None], ['b1', None], ['b2', None]] | [['a1', 1], ['b1', 2], ['b2', 2]] | Failed |
| no views | [] | [] | Passed |
| empty sources | [] | [] | Passed |
| zero label copied | [[0, None], [1, None]] | [[0, 0], [1, 0]] | Failed |
| single view | [['only', None]] | [['only', 1]] | Failed |
| source order | [['b0', None], ['a0', None]] | [['b0', 2], ['a0', 1]] | Failed |
SHA-256 / 68aded5df9701b72911f782cdbaa7e6e5687a50a6f2f9a5509c78a4c38210c7f
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(examples):
return [[v,e[1]] for e in examples for v in e[2][:1]]
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('all views inherit',solve([['s',N,['v1','v2']]]),[['v1',N],['v2',N]])
check('multiple sources',solve([['a',N,['a1']],['b',N+1,['b1','b2']]]),[['a1',N],['b1',N+1],['b2',N+1]])
check('no views',solve([['x',0,[]]]),[])
check('empty sources',solve([]),[])
check('zero label copied',solve([['x',0,[0,N]]]),[[0,0],[N,0]])
check('single view',solve([['x',N,['only']]]),[['only',N]])
check('source order',solve([['b',2,['b0']],['a',1,['a0']]]),[['b0',2],['a0',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 |
|---|---|---|---|
| all views inherit | [['v1', 1]] | [['v1', 1], ['v2', 1]] | Failed |
| multiple sources | [['a1', 1], ['b1', 2]] | [['a1', 1], ['b1', 2], ['b2', 2]] | Failed |
| no views | [] | [] | Passed |
| empty sources | [] | [] | Passed |
| zero label copied | [[0, 0]] | [[0, 0], [1, 0]] | Failed |
| single view | [['only', 1]] | [['only', 1]] | Passed |
| source order | [['b0', 2], ['a0', 1]] | [['b0', 2], ['a0', 1]] | Passed |
SHA-256 / e7450883a8ca75c94d905e4a69a5b1089fcb71be59c64eaa2ade3030b5f19f66
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.
Sign in to the archive ↗Verification & scope
View IDs need not be globally unique; rows retain source order. 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 2026-09-30T16:48:07.976625+00:00.
Case digest / 1ce78b42425c2be7a59603bef03aa14479eef54073ad3e43f9fa6ae32c30b687