FA-98296 / Autoregressive target shifting / Open access

A longer target prefix repeats the final context in extra pairs · case 01

A longer target prefix repeats the final context in extra pairs.

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

ROOT CAUSE

Pair generation repeats the last context when target positions outnumber context positions.

THE FAILURE

Pair generation repeats the last context when target positions outnumber context positions.

Unsuccessful approach: Allowing one extra lookahead pair still duplicates the final available context when targets extend further.

Case contract

Return zipped context and target pairs; unmatched suffix on either side is ignored.

Why this case matters

A small deterministic token-stream contract makes next-token alignment errors inspectable.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(context, target):
    return [[context[min(i,len(context)-1)],target[i]] for i in range(len(target)) if context]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('target longer',solve([1],[4,5]),[[1,4]])
check('context longer',solve([1,2],[4]),[[1,4]])
check('equal',solve([N,N+1],[7,8]),[[N,7],[N+1,8]])
check('empty target',solve([1],[]),[])
check('empty context',solve([],[2]),[])
check('both empty',solve([],[]),[])
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
target longer	`[[1, 4], [1, 5]]`	`[[1, 4]]`	Failed
context longer	`[[1, 4]]`	`[[1, 4]]`	Passed
equal	`[[1, 7], [2, 8]]`	`[[1, 7], [2, 8]]`	Passed
empty target	`[]`	`[]`	Passed
empty context	`[]`	`[]`	Passed
both empty	`[]`	`[]`	Passed

SHA-256 / 19e540689ba2a791c8d21a837a9bb073a1cf2c6887e4be423a604b693ff3c960

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(context, target):
    return [[context[min(i,len(context)-1)],target[i]] for i in range(min(len(target),len(context)+1))] if context else []
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('target longer',solve([1],[4,5]),[[1,4]])
check('context longer',solve([1,2],[4]),[[1,4]])
check('equal',solve([N,N+1],[7,8]),[[N,7],[N+1,8]])
check('empty target',solve([1],[]),[])
check('empty context',solve([],[2]),[])
check('both empty',solve([],[]),[])
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
target longer	`[[1, 4], [1, 5]]`	`[[1, 4]]`	Failed
context longer	`[[1, 4]]`	`[[1, 4]]`	Passed
equal	`[[1, 7], [2, 8]]`	`[[1, 7], [2, 8]]`	Passed
empty target	`[]`	`[]`	Passed
empty context	`[]`	`[]`	Passed
both empty	`[]`	`[]`	Passed

SHA-256 / baf63ba2a28ec0b100fa85fdb51fb9d9cdd6bcd1782c01b1d078133af59747f1

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

Inputs are toy integer token sequences; this contract says nothing about model quality, batching, sampling, or production training frameworks. 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 2026-09-30T16:48:11.499090+00:00.

Case digest / c72d968eafa40039232c372917f9c770fdc31fd8e076781fe585cd85ed7d6cd1