FA-100726 / Speculative decoding bookkeeping / Open access
Decoding: Mismatch report returns the draft token instead of the verified token · case 01
Mismatch report returns the draft token instead of the verified token.
ROOT CAUSE
The draft token at the mismatch position is reported as if it were verified.
THE FAILURE
The draft token at the mismatch position is reported as if it were verified.
Unsuccessful approach: Returning the first verified token fixes only rejection at position zero.
Case contract
For a nonnegative zero-based rejection index k, return verified[k] when present, otherwise None.
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, verified, k):
return draft[k] if k < len(draft) else None
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('middle mismatch', solve([1,2,3],[1,9,3],1),9)
check('first mismatch', solve([5],[8],0),8)
check('beyond verified', solve([1,2],[1],1),None)
check('empty verified', solve([1],[],0),None)
check('later position', solve([1,2,3],[1,2,7],2),7)
check('singleton token', solve([N],[N+1],0),N+1)
check('short draft but verified slot', solve([1],[1,4],1),4)
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 |
|---|---|---|---|
| middle mismatch | 2 | 9 | Failed |
| first mismatch | 5 | 8 | Failed |
| beyond verified | 2 | None | Failed |
| empty verified | 1 | None | Failed |
| later position | 3 | 7 | Failed |
| singleton token | 1 | 2 | Failed |
| short draft but verified slot | None | 4 | Failed |
SHA-256 / 48d57797d59a024da873f6f81118eb63d629e90029a8a78fcb20fb62691cc76b
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(draft, verified, k):
return verified[0] if verified else None
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('middle mismatch', solve([1,2,3],[1,9,3],1),9)
check('first mismatch', solve([5],[8],0),8)
check('beyond verified', solve([1,2],[1],1),None)
check('empty verified', solve([1],[],0),None)
check('later position', solve([1,2,3],[1,2,7],2),7)
check('singleton token', solve([N],[N+1],0),N+1)
check('short draft but verified slot', solve([1],[1,4],1),4)
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 |
|---|---|---|---|
| middle mismatch | 1 | 9 | Failed |
| first mismatch | 8 | 8 | Passed |
| beyond verified | 1 | None | Failed |
| empty verified | None | None | Passed |
| later position | 1 | 7 | Failed |
| singleton token | 2 | 2 | Passed |
| short draft but verified slot | 1 | 4 | Failed |
SHA-256 / 69d617011432d660b18541974d54df696317f5bece1b68aab3bf8fae5810d685
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 2026-09-30T16:48:39.096632+00:00.
Case digest / 3558247370b7b9aba6b6dfc37fb9f268ce4278036796da584cde0dfce9ba30a0