FA-100706 / Speculative decoding bookkeeping / Open access
Decoding: Only the final request contributes to the batch acceptance total · case 01
Only the final request contributes to the batch acceptance total.
ROOT CAUSE
Only the last request contributes to the batch total.
THE FAILURE
Only the last request contributes to the batch total.
Unsuccessful approach: Summing all proposals charges rejected suffixes.
Case contract
For rows of proposed token counts and accepted counts, sum each accepted count after bounding it to that row size.
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(proposed, accepted):
return accepted[-1] if accepted else 0
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('two requests', solve([3,2],[2,1]),3)
check('empty batch', solve([],[]),0)
check('zero accepted', solve([N],[0]),0)
check('oversized row', solve([1,3],[8,2]),3)
check('short counts', solve([2,4],[1]),1)
check('negative count', solve([3],[-1]),0)
check('distinct totals', solve([1,4],[1,3]),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 |
|---|---|---|---|
| two requests | 1 | 3 | Failed |
| empty batch | 0 | 0 | Passed |
| zero accepted | 0 | 0 | Passed |
| oversized row | 2 | 3 | Failed |
| short counts | 1 | 1 | Passed |
| negative count | -1 | 0 | Failed |
| distinct totals | 3 | 4 | Failed |
SHA-256 / c1b5a7086b8a3971c15d05aeb9c1e35832962d90a32155e6a3685796ff3da5e5
2 / The unsuccessful fix
Exit 1"""Failure Map reference implementation. Python standard library only."""
import json
N = 1
observations = []
def solve(proposed, accepted):
return sum(proposed)
def check(label, actual, expected):
observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('two requests', solve([3,2],[2,1]),3)
check('empty batch', solve([],[]),0)
check('zero accepted', solve([N],[0]),0)
check('oversized row', solve([1,3],[8,2]),3)
check('short counts', solve([2,4],[1]),1)
check('negative count', solve([3],[-1]),0)
check('distinct totals', solve([1,4],[1,3]),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 |
|---|---|---|---|
| two requests | 5 | 3 | Failed |
| empty batch | 0 | 0 | Passed |
| zero accepted | 1 | 0 | Failed |
| oversized row | 4 | 3 | Failed |
| short counts | 6 | 1 | Failed |
| negative count | 3 | 0 | Failed |
| distinct totals | 5 | 4 | Failed |
SHA-256 / a5b5a229b925bd6923275fb6b79ef6028108de36ab908f9aa76d0b3ee84afb1c
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 2026-09-30T16:48:38.598268+00:00.
Case digest / db85a7b866d4d3097869c1bd68a18d0d2e9e5606e26e8af3d01e908f787fcf3b