FA-99656 / Chunk citation alignment / Open access

Citation starts before the retained window · case 01

Citation starts before the retained window.

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

ROOT CAUSE

The clipped excerpt start is not applied to citations.

THE FAILURE

The clipped excerpt start is not applied to citations.

Unsuccessful approach: Moving an earlier start to the window boundary but retaining the old end leaks text beyond the window.

Case contract

Given window [b,b+L) and citation [s,e), return their half-open intersection or None for empty intersection.

Why this case matters

A deterministic citation bookkeeping model isolates source attribution from rendering and retrieval.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(base, span, length):
    s,e=span
    return [s,e]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('clip left and right',solve(N,[N-3,N+8],5),[N,N+5])
check('inside window',solve(N,[N+1,N+3],5),[N+1,N+3])
check('before window',solve(N,[N-4,N-1],5),None)
check('after window',solve(N,[N+6,N+8],5),None)
check('touch edge',solve(N,[N-2,N],5),None)
check('zero window',solve(N,[N,N+1],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
clip left and right	`[-2, 9]`	`[1, 6]`	Failed
inside window	`[2, 4]`	`[2, 4]`	Passed
before window	`[-3, 0]`	`None`	Failed
after window	`[7, 9]`	`None`	Failed
touch edge	`[-1, 1]`	`None`	Failed
zero window	`[1, 2]`	`None`	Failed

SHA-256 / fa86c6e327f0802417d9eaec3188a495ff541825fbb6eac53f5aaaa18c7e09f4

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(base, span, length):
    s,e=span
    return [max(s,base),e] if e>base else None
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('clip left and right',solve(N,[N-3,N+8],5),[N,N+5])
check('inside window',solve(N,[N+1,N+3],5),[N+1,N+3])
check('before window',solve(N,[N-4,N-1],5),None)
check('after window',solve(N,[N+6,N+8],5),None)
check('touch edge',solve(N,[N-2,N],5),None)
check('zero window',solve(N,[N,N+1],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
clip left and right	`[1, 9]`	`[1, 6]`	Failed
inside window	`[2, 4]`	`[2, 4]`	Passed
before window	`None`	`None`	Passed
after window	`[7, 9]`	`None`	Failed
touch edge	`None`	`None`	Passed
zero window	`[1, 2]`	`None`	Failed

SHA-256 / 1c4c83f6fee96f835d14891bbd89cb8feb18e099577e51163d47830a1a8fe50f

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 small integer offsets and identifiers under the stated toy contract; this does not define a document format or production citation API. 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 2026-09-30T16:48:26.712091+00:00.

Case digest / b10903eb1809cbf96fd32d7b8328cef9005eb6bffbdd2998302b47ac8ed1efba