FA-98721 / Model quantization arithmetic / Open access

Quantization: Dot product silently drops unmatched tail · case 01

Dot product silently drops unmatched tail.

Verified by executionVariant 1 · 6 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

Vectors must have equal lengths; return None on mismatch, otherwise compute the zero-point-centered integer dot product.

THE FAILURE

Vectors must have equal lengths; return None on mismatch, otherwise compute the zero-point-centered integer dot product.

Unsuccessful approach: The attempted repair checks the left vector length only; a short right vector still gets silently truncated by the zip-style accumulation.

Case contract

Both vectors must have exactly the declared length. Return None if either length differs; otherwise compute the zero-point-centered integer dot product.

Why this case matters

This deterministic model isolates one quantization arithmetic contract from tensor storage and runtime kernels.

1 / The failure

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(a, b, zero_a, zero_b, length):
    return sum((x-zero_a)*(y-zero_b) for x,y in zip(a,b))
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('equal declared lengths',solve([1,2],[3,4],0,0,2),11)
check('left too long',solve([1,2,3],[4,5],0,0,2),None)
check('right too long',solve([1],[2,3],0,0,1),None)
check('both short',solve([1],[2],0,0,2),None)
check('empty valid',solve([],[],N,N,0),0)
check('singleton centered',solve([N+1],[N+2],N,N,1),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
equal declared lengths	`11`	`11`	Passed
left too long	`14`	`None`	Failed
right too long	`2`	`None`	Failed
both short	`2`	`None`	Failed
empty valid	`0`	`0`	Passed
singleton centered	`2`	`2`	Passed

SHA-256 / 46008924ff0e47f439387fccb0c9128687a2e954be7c4e1bf8a3dd0015570bb5

2 / The unsuccessful fix

Exit 1

"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(a, b, zero_a, zero_b, length):
    if len(a)!=length: return None
    return sum((a[i]-zero_a)*(b[i]-zero_b) for i in range(min(len(a),len(b))))
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('equal declared lengths',solve([1,2],[3,4],0,0,2),11)
check('left too long',solve([1,2,3],[4,5],0,0,2),None)
check('right too long',solve([1],[2,3],0,0,1),None)
check('both short',solve([1],[2],0,0,2),None)
check('empty valid',solve([],[],N,N,0),0)
check('singleton centered',solve([N+1],[N+2],N,N,1),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
equal declared lengths	`11`	`11`	Passed
left too long	`None`	`None`	Passed
right too long	`2`	`None`	Failed
both short	`None`	`None`	Passed
empty valid	`0`	`0`	Passed
singleton centered	`2`	`2`	Passed

SHA-256 / 25e83f8ffb043a5b6dbc4fd59fc8c028606c0ad307ab6105123bf040880115f2

HELD IN THE MEMBER ARCHIVE

The verified repair and its recorded checks are member-only.

This mechanism has 6 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.

Verification & scope

Toy scalar/list arithmetic only; it does not model a production quantizer, hardware instruction, or accuracy/performance gain. 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 2026-09-30T16:48:16.280208+00:00.

Case digest / b4ae7c2899e2672d6a548105ce4c2cd609928ae82210532b2f86957c929cb877