FA-100696 / Speculative decoding bookkeeping / Open access
Decoding: Absolute position omits the replacement token after rejection · case 01
Absolute position omits the replacement token after rejection.
ROOT CAUSE
The replacement token is omitted from absolute-position advancement.
THE FAILURE
The replacement token is omitted from absolute-position advancement.
Unsuccessful approach: Adding a constant two advances correctly only when exactly one draft token was accepted.
Case contract
Advance absolute position by accepted draft tokens plus one replacement token after rejection.
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(position, accepted):
return position+accepted
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('partial advance', solve(10,2),13)
check('zero accepted', solve(4,0),5)
check('single accepted', solve(0,1),2)
check('negative base position', solve(-2,2),1)
check('vary position', solve(N,1),N+2)
check('large accepted', solve(3,4),8)
check('no prefix', solve(9,0),10)
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 |
|---|---|---|---|
| partial advance | 12 | 13 | Failed |
| zero accepted | 4 | 5 | Failed |
| single accepted | 1 | 2 | Failed |
| negative base position | 0 | 1 | Failed |
| vary position | 2 | 3 | Failed |
| large accepted | 7 | 8 | Failed |
| no prefix | 9 | 10 | Failed |
SHA-256 / 3aecfb9648b780d6ecda39d09ce988d59f2bf104a89d6a3ee2f58e34665301f4
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(position, accepted):
return position+2
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('partial advance', solve(10,2),13)
check('zero accepted', solve(4,0),5)
check('single accepted', solve(0,1),2)
check('negative base position', solve(-2,2),1)
check('vary position', solve(N,1),N+2)
check('large accepted', solve(3,4),8)
check('no prefix', solve(9,0),10)
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 |
|---|---|---|---|
| partial advance | 12 | 13 | Failed |
| zero accepted | 6 | 5 | Failed |
| single accepted | 2 | 2 | Passed |
| negative base position | 0 | 1 | Failed |
| vary position | 3 | 3 | Passed |
| large accepted | 5 | 8 | Failed |
| no prefix | 11 | 10 | Failed |
SHA-256 / f44cd12e6ccca251a3ce41052935c3d3cc1b588927c093cdebed10dc415a5be7
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 2026-09-30T16:48:38.598986+00:00.
Case digest / 08314f6855dc9b58196603345fa36b620fe88f4e0b5402dfba498e95392fa465