FA-98201 / Attention mask construction / Open access

Causal position does not reset at segment boundaries · case 01

Causal position does not reset at segment boundaries.

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

ROOT CAUSE

Absolute flattened index controls causality across separately indexed segments.

THE FAILURE

Absolute flattened index controls causality across separately indexed segments.

Unsuccessful approach: Resetting at segment start but ignoring segment labels permits another segment with the same local position.

Case contract

Each entry is [segment_id, local_position]; visible iff IDs match and key local position<=query local position.

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(segments):
    return [[j<=i for j in range(len(segments))] for i in range(len(segments))]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
s=[[0,i] for i in range(N)]+[[1,0],[1,1]]
check('reset and isolate',solve(s),[[s[j][0]==s[i][0] and s[j][1]<=s[i][1] for j in range(len(s))] for i in range(len(s))])
check('one segment',solve([[4,0],[4,1]]),[[True,False],[True,True]])
check('same local index different segment',solve([[1,0],[2,0]]),[[True,False],[False,True]])
check('empty',solve([]),[])
check('singleton',solve([[3,7]]),[[True]])
check('local order differs global',solve([[5,1],[5,4]]),[[True,False],[True,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
reset and isolate	`[[True, False, False], [True, True, False], [True, True, True]]`	`[[True, False, False], [False, True, False], [False, True, True]]`	Failed
one segment	`[[True, False], [True, True]]`	`[[True, False], [True, True]]`	Passed
same local index different segment	`[[True, False], [True, True]]`	`[[True, False], [False, True]]`	Failed
empty	`[]`	`[]`	Passed
singleton	`[[True]]`	`[[True]]`	Passed
local order differs global	`[[True, False], [True, True]]`	`[[True, False], [True, True]]`	Passed

SHA-256 / 5ed60c5c6d000ea06ff015eb261dd1d3555fab36aa297c3487df74842429a9d5

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(segments):
    return [[segments[j][1]<=segments[i][1] for j in range(len(segments))] for i in range(len(segments))]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
s=[[0,i] for i in range(N)]+[[1,0],[1,1]]
check('reset and isolate',solve(s),[[s[j][0]==s[i][0] and s[j][1]<=s[i][1] for j in range(len(s))] for i in range(len(s))])
check('one segment',solve([[4,0],[4,1]]),[[True,False],[True,True]])
check('same local index different segment',solve([[1,0],[2,0]]),[[True,False],[False,True]])
check('empty',solve([]),[])
check('singleton',solve([[3,7]]),[[True]])
check('local order differs global',solve([[5,1],[5,4]]),[[True,False],[True,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
reset and isolate	`[[True, True, False], [True, True, False], [True, True, True]]`	`[[True, False, False], [False, True, False], [False, True, True]]`	Failed
one segment	`[[True, False], [True, True]]`	`[[True, False], [True, True]]`	Passed
same local index different segment	`[[True, True], [True, True]]`	`[[True, False], [False, True]]`	Failed
empty	`[]`	`[]`	Passed
singleton	`[[True]]`	`[[True]]`	Passed
local order differs global	`[[True, False], [True, True]]`	`[[True, False], [True, True]]`	Passed

SHA-256 / b122c3a9ee3cc8f748fded585694e0a6ed172bff1ae0b003fe3f5e282467e347

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 2026-09-30T16:48:10.629822+00:00.

Case digest / 3199df36c851be76c94a10fc2afe6fb9a7e0cc0883a8fb3d149cb826a0c47048