FA-100606 / Mixture of experts dispatch / Open access

Router scores are paired with the wrong expert IDs · case 01

Router scores are paired with the wrong expert IDs.

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

ROOT CAUSE

Sorting the expert names independently detaches them from their score-vector positions.

THE FAILURE

Sorting the expert names independently detaches them from their score-vector positions.

Unsuccessful approach: Sorting the scores preserves their ranking but still corrupts the name-to-score association.

Case contract

Given expert_names and score vector of equal length, return (name,score) pairs in input index order.

Why this case matters

A deterministic toy router makes this dispatch decision independently inspectable.

1 / The failure

Exit 1

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

N = 1
observations = []
def solve(expert_names,scores):
    return list(zip(sorted(expert_names),scores))
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
names=['z'+str(N),'a'+str(N),'m'+str(N)]; vals=[N,2*N,3*N]
check('index association',solve(names,vals),[[names[0],N],[names[1],2*N],[names[2],3*N]])
check('singleton',solve(['x'],[N]),[['x',N]])
check('empty',solve([],[]),[])
check('zero scores',solve(['b','a'],[0,0]),[['b',0],['a',0]])
check('negative',solve(['x','y'],[-N,N]),[['x',-N],['y',N]])
check('order retained',solve(['c','a','b'],[3,1,2]),[['c',3],['a',1],['b',2]])
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
index association	`[['a1', 1], ['m1', 2], ['z1', 3]]`	`[['z1', 1], ['a1', 2], ['m1', 3]]`	Failed
singleton	`[['x', 1]]`	`[['x', 1]]`	Failed
empty	`[]`	`[]`	Passed
zero scores	`[['a', 0], ['b', 0]]`	`[['b', 0], ['a', 0]]`	Failed
negative	`[['x', -1], ['y', 1]]`	`[['x', -1], ['y', 1]]`	Failed
order retained	`[['a', 3], ['b', 1], ['c', 2]]`	`[['c', 3], ['a', 1], ['b', 2]]`	Failed

SHA-256 / 86958edff86a77d13d64c5da39c709dbcafd0da338cc6f2524b06dbabe419b82

2 / The unsuccessful fix

Exit 1

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

N = 1
observations = []
def solve(expert_names,scores):
    return list(zip(expert_names,sorted(scores,reverse=True)))
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
names=['z'+str(N),'a'+str(N),'m'+str(N)]; vals=[N,2*N,3*N]
check('index association',solve(names,vals),[[names[0],N],[names[1],2*N],[names[2],3*N]])
check('singleton',solve(['x'],[N]),[['x',N]])
check('empty',solve([],[]),[])
check('zero scores',solve(['b','a'],[0,0]),[['b',0],['a',0]])
check('negative',solve(['x','y'],[-N,N]),[['x',-N],['y',N]])
check('order retained',solve(['c','a','b'],[3,1,2]),[['c',3],['a',1],['b',2]])
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
index association	`[['z1', 3], ['a1', 2], ['m1', 1]]`	`[['z1', 1], ['a1', 2], ['m1', 3]]`	Failed
singleton	`[['x', 1]]`	`[['x', 1]]`	Failed
empty	`[]`	`[]`	Passed
zero scores	`[['b', 0], ['a', 0]]`	`[['b', 0], ['a', 0]]`	Failed
negative	`[['x', 1], ['y', -1]]`	`[['x', -1], ['y', 1]]`	Failed
order retained	`[['c', 3], ['a', 2], ['b', 1]]`	`[['c', 3], ['a', 1], ['b', 2]]`	Failed

SHA-256 / a1fcce2f139756352efb0da124ff2eab18b46614fa73847547c7a2c2dbb500d4

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 finite toy scores and routes; the contract does not model a production model, accelerator, or learned routing quality. 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 2026-09-30T16:48:37.646403+00:00.

Case digest / 18affddfd481e5ae4958b79a688953060fa64fcd134515bffaa58dbbef16bee6