FA-100486 / Sparse tensor canonicalization / Open access

Sparse tensor: COO matrix-vector product ignores row coordinates · case 01

COO matrix-vector product ignores row coordinates.

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

ROOT CAUSE

COO entries require both coordinates to select the output row and input-vector element.

THE FAILURE

COO entries require both coordinates to select the output row and input-vector element.

Unsuccessful approach: An inclusive row prefix still mixes contributions from multiple rows rather than computing each row’s own sum.

Case contract

Return rows-length integer product y where each [r,c,v] contributes v*vector[c] to y[r].

Why this case matters

A small offline sparse-array operation has explicit coordinate and storage semantics.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(entries,vector,rows):
    return [sum(v*vector[c] for r,c,v in entries)]*rows
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('separate rows',solve([[0,1,N],[2,0,3]],[4,5],3),[5*N,0,12])
check('empty matrix',solve([],[1,2],N),[0]*N)
check('single row',solve([[0,0,7]],[N],1),[7*N])
check('same row add',solve([[1,0,2],[1,1,3]],[N,4],3),[0,2*N+12,0])
check('zero value',solve([[0,0,0]],[N],1),[0])
check('trailing empty rows',solve([[0,0,1]],[N],3),[N,0,0])
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
separate rows	`[17, 17, 17]`	`[5, 0, 12]`	Failed
empty matrix	`[0]`	`[0]`	Passed
single row	`[7]`	`[7]`	Passed
same row add	`[14, 14, 14]`	`[0, 14, 0]`	Failed
zero value	`[0]`	`[0]`	Passed
trailing empty rows	`[1, 1, 1]`	`[1, 0, 0]`	Failed

SHA-256 / d6c553b55646a255816d3ee179ce29cf0d470b759702b6b2b2c2d7086dd9a163

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(entries,vector,rows):
    return [sum(v*vector[c] for r,c,v in entries if r<=i) for i in range(rows)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('separate rows',solve([[0,1,N],[2,0,3]],[4,5],3),[5*N,0,12])
check('empty matrix',solve([],[1,2],N),[0]*N)
check('single row',solve([[0,0,7]],[N],1),[7*N])
check('same row add',solve([[1,0,2],[1,1,3]],[N,4],3),[0,2*N+12,0])
check('zero value',solve([[0,0,0]],[N],1),[0])
check('trailing empty rows',solve([[0,0,1]],[N],3),[N,0,0])
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
separate rows	`[5, 5, 17]`	`[5, 0, 12]`	Failed
empty matrix	`[0]`	`[0]`	Passed
single row	`[7]`	`[7]`	Passed
same row add	`[0, 14, 14]`	`[0, 14, 0]`	Failed
zero value	`[0]`	`[0]`	Passed
trailing empty rows	`[1, 1, 1]`	`[1, 0, 0]`	Failed

SHA-256 / cc50cc5ba7a066664e58c597adc9edb76eeb9286b0d5210fb8212576bacd082a

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

These are toy integer-valued COO or CSR contracts; they make no claim about a framework API, floating point policy, or large tensor performance. 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 2026-09-30T16:48:36.272465+00:00.

Case digest / f3b361cfb70b78e156be25c974f89f568e734d23f2d60367aa6df2385d5229a4