FA-100686 / Speculative decoding bookkeeping / Open access
Decoding: Target evaluation count omits the rejected token comparison · case 01
Target evaluation count omits the rejected token comparison.
ROOT CAUSE
The check at the rejecting position is excluded from work accounting.
THE FAILURE
The check at the rejecting position is excluded from work accounting.
Unsuccessful approach: Always charging one extra evaluation overcounts fully accepted proposals.
Case contract
If rejection is at zero-based index k, target evaluations equal k+1; if fully accepted, they equal the draft length.
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(draft, rejected_at):
return rejected_at if rejected_at is not None else len(draft)
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('first rejection', solve([1,2],0),1)
check('later rejection', solve([1,2,3],2),3)
check('full verification', solve([1,2],None),2)
check('empty verified batch', solve([],None),0)
check('last rejection', solve([1,2],1),2)
check('singleton rejection', solve([N],0),1)
check('rejection at penultimate slot', solve([1,2,3,4],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 |
|---|---|---|---|
| first rejection | 0 | 1 | Failed |
| later rejection | 2 | 3 | Failed |
| full verification | 2 | 2 | Passed |
| empty verified batch | 0 | 0 | Passed |
| last rejection | 1 | 2 | Failed |
| singleton rejection | 0 | 1 | Failed |
| rejection at penultimate slot | 2 | 3 | Failed |
SHA-256 / a00b560a4a49185575b40def21c396d1810e1dffe5dba7bfbb4985dc77926091
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(draft, rejected_at):
return len(draft)+1
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('first rejection', solve([1,2],0),1)
check('later rejection', solve([1,2,3],2),3)
check('full verification', solve([1,2],None),2)
check('empty verified batch', solve([],None),0)
check('last rejection', solve([1,2],1),2)
check('singleton rejection', solve([N],0),1)
check('rejection at penultimate slot', solve([1,2,3,4],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 |
|---|---|---|---|
| first rejection | 3 | 1 | Failed |
| later rejection | 4 | 3 | Failed |
| full verification | 3 | 2 | Failed |
| empty verified batch | 1 | 0 | Failed |
| last rejection | 3 | 2 | Failed |
| singleton rejection | 2 | 1 | Failed |
| rejection at penultimate slot | 5 | 3 | Failed |
SHA-256 / c6454338206eef0313d896f4546b51bb8cb6d7a12c16318c803217a4b9c1ca12
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 2026-09-30T16:48:38.598891+00:00.
Case digest / 454898f9237a5b2205e34c859e6eec66cf9551b69481e9bdd4d7e8eb598d8f11