FA-99461 / Vector retrieval ranking / Open access

Vector retrieval: Candidates exactly on the score floor disappear · case 01

Candidates exactly on the score floor disappear.

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

ROOT CAUSE

The cutoff uses a strict comparison.

THE FAILURE

The cutoff uses a strict comparison.

Unsuccessful approach: An epsilon offset still excludes values exactly equal to the floor.

Case contract

Keep rows [id,score] when score is at least floor, preserving input order.

Why this case matters

An offline vector retrieval toy model isolates one ranking contract from embedding generation, storage, and serving.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(rows, floor):
    return [r[0] for r in rows if r[1]>floor]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('exact boundary',solve([['edge',N],['above',N+1]],N),['edge','above'])
check('boundary only',solve([['edge',N]],N),['edge'])
check('negative floor',solve([['edge',-N],['low',-N-1]],-N),['edge'])
check('fraction boundary',solve([['edge',1.25],['above',1.5]],1.25),['edge','above'])
check('empty',solve([],N),[])
check('all above',solve([['x',N+2],['y',N+3]],N),['x','y'])
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
exact boundary	`['above']`	`['edge', 'above']`	Failed
boundary only	`[]`	`['edge']`	Failed
negative floor	`[]`	`['edge']`	Failed
fraction boundary	`['above']`	`['edge', 'above']`	Failed
empty	`[]`	`[]`	Passed
all above	`['x', 'y']`	`['x', 'y']`	Passed

SHA-256 / 795145ba7191a1404815d88ec6a382e9c66e66f66b8a97c24d0a7de92a0dcbaa

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(rows, floor):
    return [r[0] for r in rows if r[1]>=floor+1e-9]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
check('exact boundary',solve([['edge',N],['above',N+1]],N),['edge','above'])
check('boundary only',solve([['edge',N]],N),['edge'])
check('negative floor',solve([['edge',-N],['low',-N-1]],-N),['edge'])
check('fraction boundary',solve([['edge',1.25],['above',1.5]],1.25),['edge','above'])
check('empty',solve([],N),[])
check('all above',solve([['x',N+2],['y',N+3]],N),['x','y'])
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
exact boundary	`['above']`	`['edge', 'above']`	Failed
boundary only	`[]`	`['edge']`	Failed
negative floor	`[]`	`['edge']`	Failed
fraction boundary	`['above']`	`['edge', 'above']`	Failed
empty	`[]`	`[]`	Passed
all above	`['x', 'y']`	`['x', 'y']`	Passed

SHA-256 / 2656413ba73515bd5327c295a1384038c51b84939a069525eed585ac2dd3c333

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

Inputs are small exact toy vectors and scores. No production index, ANN recall, embedding quality, or latency claim is implied. 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 2026-09-30T16:48:24.474761+00:00.

Case digest / 965cb3ac704d9839d87781a7804e56d51fafb41f6f4510e512e05b134ce4915e