FA-100716 / Speculative decoding bookkeeping / Open access
Decoding: Flags after the first rejection are counted as accepted · case 01
Flags after the first rejection are counted as accepted.
ROOT CAUSE
True flags after a rejected position are counted as accepted.
THE FAILURE
True flags after a rejected position are counted as accepted.
Unsuccessful approach: Requiring all flags to be true returns zero for a partially accepted prefix.
Case contract
Return the number of consecutive true flags from the beginning.
Why this case matters
A self-contained bookkeeping model makes prefix and token-count semantics observable without a model runtime.
1 / The failure
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(flags):
return sum(bool(v) for v in flags)
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('hole stops', solve([True,False,True]),1)
check('all accepted', solve([True]*N),N)
check('reject first', solve([False,True]),0)
check('empty flags', solve([]),0)
check('singleton true', solve([True]),1)
check('singleton false', solve([False]),0)
check('later false', solve([True,True,False,True]),2)
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 |
|---|---|---|---|
| hole stops | 2 | 1 | Failed |
| all accepted | 1 | 1 | Passed |
| reject first | 1 | 0 | Failed |
| empty flags | 0 | 0 | Passed |
| singleton true | 1 | 1 | Passed |
| singleton false | 0 | 0 | Passed |
| later false | 3 | 2 | Failed |
SHA-256 / a9922b5aee5e588a0a08ead8e5417ae02d323f6ca4c6153650ae3e4d55d35115
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(flags):
return len(flags) if all(flags) else 0
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('hole stops', solve([True,False,True]),1)
check('all accepted', solve([True]*N),N)
check('reject first', solve([False,True]),0)
check('empty flags', solve([]),0)
check('singleton true', solve([True]),1)
check('singleton false', solve([False]),0)
check('later false', solve([True,True,False,True]),2)
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 |
|---|---|---|---|
| hole stops | 0 | 1 | Failed |
| all accepted | 1 | 1 | Passed |
| reject first | 0 | 0 | Passed |
| empty flags | 0 | 0 | Passed |
| singleton true | 1 | 1 | Passed |
| singleton false | 0 | 0 | Passed |
| later false | 0 | 2 | Failed |
SHA-256 / 5d5631ed2c277e2dbca63c895b678a26ab6c04d8d1e3d21ae082a03f4356bb9b
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
These are finite toy sequences with explicit contracts; they make no claim about production decoder behavior or model quality. 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 2026-09-30T16:48:38.598007+00:00.
Case digest / 50fa91b93df15c45f27f90f249399728196a5927e4cbe9e30e3eb71a49fe7048