FA-98076 / Sequence batch packing / Open access

Overlapping windows advance by the overlap instead of stride · case 01

Overlapping windows advance by the overlap instead of stride.

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

ROOT CAUSE

The start increment is set to overlap length, causing excessive repetition.

THE FAILURE

The start increment is set to overlap length, causing excessive repetition.

Unsuccessful approach: Adding one to the stride repairs some windows but skips a valid start for larger widths.

Case contract

Emit windows of width, advancing by width-overlap; stop after the window reaches the sequence end. A final shorter window is retained.

Why this case matters

This deterministic toy packing rule models a common boundary in preparing bounded training sequences.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(tokens, width, overlap):
    return [tokens[i:i+width] for i in range(0,len(tokens),max(1,overlap))]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
data=list(range(2*N+1)); w=N+1; o=1
check('overlap and tail', solve(data,w,o), [data[:w],data[N:2*N+1]])
tokens=list(range(N+2))
check('no overlap', solve(tokens,N,0), [tokens[i:i+N] for i in range(0,len(tokens),N)])
check('full single window', solve([N],N+2,1), [[N]])
check('empty', solve([],N,min(1,N-1)), [])
check('overlap repeats', solve(list(range(5)),3,2), [[0,1,2],[1,2,3],[2,3,4]])
data=list(range(2*N))
expected=[list(range(N)),list(range(1,2))] if N==1 else [list(range(N)),list(range(N-1,2*N-1)),list(range(2*N-2,2*N))]
check('exact end', solve(data,N,min(1,N-1)), expected)
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
overlap and tail	`[[0, 1], [1, 2], [2]]`	`[[0, 1], [1, 2]]`	Failed
no overlap	`[[0], [1], [2]]`	`[[0], [1], [2]]`	Passed
full single window	`[[1]]`	`[[1]]`	Passed
empty	`[]`	`[]`	Passed
overlap repeats	`[[0, 1, 2], [2, 3, 4], [4]]`	`[[0, 1, 2], [1, 2, 3], [2, 3, 4]]`	Failed
exact end	`[[0], [1]]`	`[[0], [1]]`	Passed

SHA-256 / bc1e5648b24efa070d7da8226fa3b9ae95079db0b8e0c4963a50657b2c3c8891

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(tokens, width, overlap):
    return [tokens[i:i+width] for i in range(0,len(tokens),width-overlap+1)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
data=list(range(2*N+1)); w=N+1; o=1
check('overlap and tail', solve(data,w,o), [data[:w],data[N:2*N+1]])
tokens=list(range(N+2))
check('no overlap', solve(tokens,N,0), [tokens[i:i+N] for i in range(0,len(tokens),N)])
check('full single window', solve([N],N+2,1), [[N]])
check('empty', solve([],N,min(1,N-1)), [])
check('overlap repeats', solve(list(range(5)),3,2), [[0,1,2],[1,2,3],[2,3,4]])
data=list(range(2*N))
expected=[list(range(N)),list(range(1,2))] if N==1 else [list(range(N)),list(range(N-1,2*N-1)),list(range(2*N-2,2*N))]
check('exact end', solve(data,N,min(1,N-1)), expected)
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
overlap and tail	`[[0, 1], [2]]`	`[[0, 1], [1, 2]]`	Failed
no overlap	`[[0], [2]]`	`[[0], [1], [2]]`	Failed
full single window	`[[1]]`	`[[1]]`	Passed
empty	`[]`	`[]`	Passed
overlap repeats	`[[0, 1, 2], [2, 3, 4], [4]]`	`[[0, 1, 2], [1, 2, 3], [2, 3, 4]]`	Failed
exact end	`[[0]]`	`[[0], [1]]`	Failed

SHA-256 / a0473dc77a211624ce56fe9d711d9136d68569f490593b74e301d73ab627915f

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 short integer-token lists under the stated toy contract; results do not claim compatibility with a specific training framework. 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 2026-09-30T16:48:08.896898+00:00.

Case digest / 4c46a75088fda22f8e0fbbddb9aa9b6221ed6b52ac68fd9602c124c1310a5e7f