FA-98191 / Attention mask construction / Open access

Dilated attention includes every intermediate key · case 01

Dilated attention includes every intermediate key.

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

ROOT CAUSE

The implementation applies a range window instead of the required stride.

THE FAILURE

The implementation applies a range window instead of the required stride.

Unsuccessful approach: Checking divisibility without the causal condition admits future keys.

Case contract

Query i sees key j iff j<=i and (i-j) is divisible by positive dilation D.

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(length, dilation):
    return [[0<=i-j<dilation for j in range(length)] for i in range(length)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('stride two',solve(N+3,2),[[j<=i and (i-j)%2==0 for j in range(N+3)] for i in range(N+3)])
check('unit dilation causal',solve(3,1),[[j<=i for j in range(3)] for i in range(3)])
check('empty',solve(0,2),[])
check('singleton',solve(1,4),[[True]])
check('future forbidden',solve(3,2)[0],[True,False,False])
check('far aligned retained',solve(5,2)[4],[True,False,True,False,True])
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
stride two	`[[True, False, False, False], [True, True, False, False], [False, True, True, False], [False, False, True, True]]`	`[[True, False, False, False], [False, True, False, False], [True, False, True, False], [False, True, False, True]]`	Failed
unit dilation causal	`[[True, False, False], [False, True, False], [False, False, True]]`	`[[True, False, False], [True, True, False], [True, True, True]]`	Failed
empty	`[]`	`[]`	Passed
singleton	`[[True]]`	`[[True]]`	Passed
future forbidden	`[True, False, False]`	`[True, False, False]`	Passed
far aligned retained	`[False, False, False, True, True]`	`[True, False, True, False, True]`	Failed

SHA-256 / 8ca8466963cb3424c2859e77df485aa978d1ebd97102b7c11d2b0b587f55b4a8

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(length, dilation):
    return [[(i-j)%dilation==0 for j in range(length)] for i in range(length)]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('stride two',solve(N+3,2),[[j<=i and (i-j)%2==0 for j in range(N+3)] for i in range(N+3)])
check('unit dilation causal',solve(3,1),[[j<=i for j in range(3)] for i in range(3)])
check('empty',solve(0,2),[])
check('singleton',solve(1,4),[[True]])
check('future forbidden',solve(3,2)[0],[True,False,False])
check('far aligned retained',solve(5,2)[4],[True,False,True,False,True])
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
stride two	`[[True, False, True, False], [False, True, False, True], [True, False, True, False], [False, True, False, True]]`	`[[True, False, False, False], [False, True, False, False], [True, False, True, False], [False, True, False, True]]`	Failed
unit dilation causal	`[[True, True, True], [True, True, True], [True, True, True]]`	`[[True, False, False], [True, True, False], [True, True, True]]`	Failed
empty	`[]`	`[]`	Passed
singleton	`[[True]]`	`[[True]]`	Passed
future forbidden	`[True, False, True]`	`[True, False, False]`	Failed
far aligned retained	`[True, False, True, False, True]`	`[True, False, True, False, True]`	Passed

SHA-256 / 6778a708a0c8ce4956c47b6e547992281ceb7a9a0c57f7cd757e966b5e3dfa96

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 2026-09-30T16:48:10.193160+00:00.

Case digest / 184e60c3b63ab1cc912bcf02a9b2c890c449830bf58687a0d170857ef7018f5a