FA-98321 / Autoregressive target shifting / Open access

A boundary input predicts the first token of the next segment · case 01

A boundary input predicts the first token of the next segment.

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

ROOT CAUSE

Adjacent pairs are generated across a declared segment boundary.

THE FAILURE

Adjacent pairs are generated across a declared segment boundary.

Unsuccessful approach: Dropping every pair touching a boundary also discards valid within-segment predictions.

Case contract

Given tokens and [start,end] segments, return adjacent pairs entirely within each segment; malformed empty segments contribute none.

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(tokens, segments):
    return [[tokens[i],tokens[i+1]] for i in range(max(0,segments[0][0]),min(segments[-1][1]-1,len(tokens)-1))] if segments else []
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('two segments',solve([0,1,2,3,4],[[0,2],[2,5]]),[[0,1],[2,3],[3,4]])
check('single segment',solve([1,2,3],[[0,3]]),[[1,2],[2,3]])
check('singleton segment',solve([1,2],[[0,1],[1,2]]),[])
check('empty segments',solve([1],[]),[])
check('clipped end',solve([1,2],[[0,8]]),[[1,2]])
check('offset segment',solve([0,1,2,3],[[2,4]]),[[2,3]])
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
two segments	`[[0, 1], [1, 2], [2, 3], [3, 4]]`	`[[0, 1], [2, 3], [3, 4]]`	Failed
single segment	`[[1, 2], [2, 3]]`	`[[1, 2], [2, 3]]`	Passed
singleton segment	`[[1, 2]]`	`[]`	Failed
empty segments	`[]`	`[]`	Passed
clipped end	`[[1, 2]]`	`[[1, 2]]`	Passed
offset segment	`[[2, 3]]`	`[[2, 3]]`	Passed

SHA-256 / 904a74569f53c16b5b68e3c25b727a5e176bb05542cf1350e5ec51596444c0ee

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(tokens, segments):
    return [[tokens[i],tokens[i+1]] for a,b in segments for i in range(max(0,a+1),min(b-1,len(tokens)-1))]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('two segments',solve([0,1,2,3,4],[[0,2],[2,5]]),[[0,1],[2,3],[3,4]])
check('single segment',solve([1,2,3],[[0,3]]),[[1,2],[2,3]])
check('singleton segment',solve([1,2],[[0,1],[1,2]]),[])
check('empty segments',solve([1],[]),[])
check('clipped end',solve([1,2],[[0,8]]),[[1,2]])
check('offset segment',solve([0,1,2,3],[[2,4]]),[[2,3]])
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
two segments	`[[3, 4]]`	`[[0, 1], [2, 3], [3, 4]]`	Failed
single segment	`[[2, 3]]`	`[[1, 2], [2, 3]]`	Failed
singleton segment	`[]`	`[]`	Passed
empty segments	`[]`	`[]`	Passed
clipped end	`[]`	`[[1, 2]]`	Failed
offset segment	`[]`	`[[2, 3]]`	Failed

SHA-256 / 15ade1bef415efc9d5f0e7b9281463e7c64c273e60097e5f4127214ef43babfc

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 2026-09-30T16:48:11.933724+00:00.

Case digest / 90f72d32e6799b7fa9a03473f99289c12e386cba2622eb295318e73b734d1e02