FA-98146 / Attention mask construction / Open access

Padded keys remain visible to active queries · case 01

Padded keys remain visible to active queries.

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

ROOT CAUSE

Key validity is not included in the attention predicate.

THE FAILURE

Key validity is not included in the attention predicate.

Unsuccessful approach: Masking invalid queries alone leaves padded key columns visible.

Case contract

Return rows for every query; a key is visible iff valid and key_index<=query_index.

Why this case matters

A small explicit visibility matrix makes sequence-attention boundary behavior reviewable.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(valid):
    return [[j<=i for j in range(len(valid))] for i in range(len(valid))]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
v=[True]*N+[False]
check('padded final key',solve(v),[[j<=i and j<N for j in range(N+1)] for i in range(N+1)])
check('all valid',solve([True,True]),[[True,False],[True,True]])
check('all padded',solve([False,False]),[[False,False],[False,False]])
check('empty',solve([]),[])
check('singleton active',solve([True]),[[True]])
check('singleton pad',solve([False]),[[False]])
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
padded final key	`[[True, False], [True, True]]`	`[[True, False], [True, False]]`	Failed
all valid	`[[True, False], [True, True]]`	`[[True, False], [True, True]]`	Passed
all padded	`[[True, False], [True, True]]`	`[[False, False], [False, False]]`	Failed
empty	`[]`	`[]`	Passed
singleton active	`[[True]]`	`[[True]]`	Passed
singleton pad	`[[True]]`	`[[False]]`	Failed

SHA-256 / a0dfb9ae6ff09e3d7d24055dd4082e44e770dbb377e12b7b4d8b7f215f874847

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(valid):
    return [[bool(valid[i]) and j<=i for j in range(len(valid))] for i in range(len(valid))]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
v=[True]*N+[False]
check('padded final key',solve(v),[[j<=i and j<N for j in range(N+1)] for i in range(N+1)])
check('all valid',solve([True,True]),[[True,False],[True,True]])
check('all padded',solve([False,False]),[[False,False],[False,False]])
check('empty',solve([]),[])
check('singleton active',solve([True]),[[True]])
check('singleton pad',solve([False]),[[False]])
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
padded final key	`[[True, False], [False, False]]`	`[[True, False], [True, False]]`	Failed
all valid	`[[True, False], [True, True]]`	`[[True, False], [True, True]]`	Passed
all padded	`[[False, False], [False, False]]`	`[[False, False], [False, False]]`	Passed
empty	`[]`	`[]`	Passed
singleton active	`[[True]]`	`[[True]]`	Passed
singleton pad	`[[False]]`	`[[False]]`	Passed

SHA-256 / 415e28534c079b7c112463d593bd1d18715372438cafe25571fd20cd79608f12

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 Boolean matrix contracts; they do not define a production model architecture or tensor library behavior. 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 2026-09-30T16:48:09.758011+00:00.

Case digest / 6d48bcc429414018762769eca9b4d412e989b31797fa72fb1a9f0a1ad05b1e37